" WW7

an;
(A

w:
.ﬁaﬂr

J .30.. .1.

.1‘3’! 7”?!
a» A! 3.3.01

Lad .. .

1:: w
, x.,.r.§txz§z.:

; 1%....
L I 1..

. V... :-
..2 a «him an“. . v .
.. 1...»?
u .N’ I

i}?

~45

Sum

7.5.3.... .
V 2:1

L. .t

 

 

 

‘0.
5......»

115...”. n

_...$.gw.»....ﬁ...r..c,m%

hi 5‘

3 ..
xtJM. . iﬁ.
‘ a,“ trﬁﬁVead.

$3.333“? .. k, T
. _ $93.! 3. . . 5.3:: BR
anus», 9... mg, Edvaﬁvé... .1
s 5.2.: {15 2.43? h a...
2... amid. “33914.... Hr R.
.i ,

3;??- 142.3."
1.533!!!

a
.. .. .
inrﬁnnmm

z ‘

. 7"
I: n Av:
V. 2
“0" 1: ll
. Kﬁﬁﬂumu

 

 

 

LIBRARY
Michigan State
University

 

 

 

This is to certify that the
dissertation entitled

AN ANALYSIS OF INCOME POVERTY EFFECTS IN CASH
CROPPING ECONOMIES IN RURAL MOZAMBIQUE :
BLENDING ECONOMETRIC AND
ECONOMY-WIDE MODELS

presented by

RUI M. S. BENFICA

has been accepted towards fulﬁllment
of the requirements for the

Doctoral degree in Agricultural Economics

 

 

§
/
. , .
// / (m4? [Eggs/(Clix

C Méjd‘r‘Professor’s ignature

7 /Z‘zn/ 2’6

Date

  

 

MSU is an Afﬁrmative Action/Equal Opportunity Institution

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE

DATE DUE

DATE DUE

 

WWW;

 

1.1

I

 

0 80M 0 é”5th

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/CIRC/Date0ue.indd-p.1

 

 

AN ANALYSIS OF INCOME POVERTY EFFECTS IN CASH CROPPING
ECONOMIES IN RURAL MOZAMBIQUE: BLENDING
ECONOMETRIC AND ECONOMY-WIDE MODELS

By

Rui M.S. Benfica

A DISSERTATION

Submitted to
Michigan State University
In partial fulﬁllment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Agricultural Economics

2006

ABSTRACT
AN ANALYSIS OF INCOME POVERTY EFFECTS IN CASH CROPPING
ECONOMIES IN RURAL MOZAMBIQUE: BLENDING
ECONOMETRIC AND ECONOMY-WIDE MODELS
By
Rui M.S. Benﬁca

Contract farming is a pervasive institutional arrangement in cash cropping
economies in Mozambique. Empirical evidence on its nature and, especially, the extent to
which policies can generate broad based income growth and poverty reduction is lacking.
This Study investigates the rationale for persistence, the determinants of farmer
participation and performance in cotton and tobacco schemes (Essay One), and the
economy-wide effects of expansion and shocks in cotton and tobacco sectors on poverty
reduction in concession areas of the Zambezi valley of Mozambique (Essay Two).

In the ﬁrst essay, we ﬁnd that in both sectors contract farming is an institutional
response to widespread failure in input, credit and output markets and the absence of a
functional public and market based service provision network. Two stage econometric
procedures (testing for the existence of threshold effects in land holdings and educational
attainment) indicate that in both areas participation in the schemes is driven by factor
endowments, asset ownership and alternative income opportunities, and very little by
demographic factors. Also, there are no returns to education in either sector; this result is
consistent with previous research in Mozambique but surprising in an agronomically
demanding crop like tobacco. Farm level proﬁtability in cotton is signiﬁcantly lower than
in tobacco. Land holdings have a signiﬁcant effect on proﬁts for both crops at the highest

threshold level, but effects on total crop income and total household income are found

only in tobacco growing areas, where tobacco farmers appear not be giving up proﬁtable
off-farm Opportunities. In those areas, we ﬁnd that results may be driven by the relatively
more efﬁcient use of hired labor; labor supply in those areas is predominantly provided
by non-growers that end up sharing the beneﬁts of contract farming. Lower proﬁtability
in cotton areas is a result of low producer prices, high input costs, and lack of effective
coordination which results in low productivity and poor quality of farm output.

In the second essay, we ﬁnd that poverty reduction effects of scheme expansion
and Shocks are sizable in both areas, more so in tobacco growing areas where economic
linkages are stronger. While in tobacco areas expansion with higher export prices yields
higher beneﬁts, in cotton areas, where levels of productivity are extremely low,
expansion with productivity gains has a more broad-based effect; even when impacts are
limited among growers, any expansion in cotton production results in some beneﬁts to
non-growers. The damages of increased input prices are more severe in tobacco growing
areas, where the input package is substantially more expensive. The effects of an export
tax are more severe in tobacco growing areas where it signiﬁcantly limits the effects of
otherwise successful expansion efforts. In both areas, better maize prices have very
positive implications for poverty reduction.

The study recommends that government not embark in restrictive trade policies
(export taxes and maize export restrictions). Instead, it should promote an environment
conducive for private sector investment and improved sector coordination. Increased
contribution of cotton to rural livelihoods will require increased productivity through an
improved input package, extension, and prices to farmers. For long term sustainability in

tobacco, adverse environmental impacts deserve more attention.

DEDICATION

I dedicate this work to my wife, Marcia, and my son Ruy Allan, my
sources of joy and inspiration; and my parents for their love and
dedication!

iv

ACKNOWLEDGEMENTS

I would like to acknowledge several people and institutions that, in many different
ways, made possible the completion of my Ph.D. Program. I ought a very special thanks
to Dr. David L. Tschirley, my major professor and Dissertation Supervisor, who
encouraged me to pursue this dream, and always gave wise advice in multiple aspects of
this journey. Dr. Tschirley was instrumental in the research and Dissertation design and
completion; without his unconditional commitment this journey would not be completed.
Many thanks are also due to Dr. Duncan H. Boughton, who served in the committee, and
was instMental for the design and implementation of the ﬁeld work in Mozambique;
Dr. Boughton’s willingness to share views, ﬁnd common ground, and overcome
sometimes difﬁcult circumstances is appreciated. I thank Dr. Channing Amdt, from
Purdue University, for accepting to participate in the committee and for his outstanding
mentoring role and support provided while I was in Mozambique and throughout the
process particularly with respect to the application and understanding of economy-wide
modeling techniques. I also thank Dr. Charles Ballard, from the Economics Department
at MSU, for participating in the committee and for his practical and pragmatic
contributions.

Through Dr. Amdt’s networking, I interacted with economy-wide modeling
practitioners around the world with whom I learned a great deal. In addition to Dr. Amdt,
I recognize the important contributions of the instructors in the CGE Training Courses at
the University of Cape Town (UCT, South Africa): Sherman Robison (University of
Sussex and IDS, UK); Lawrence Edwards (U CT), Dirk van Seventer (TIPS); James

Thurow (University of Natal and IFPRI); Rob Davis; Scott McDonald (University of

Sheffield, UK); and Joe Francois (Erasmus University, The Netherlands). A Special
appreciation goes to Professor Sherman Robinson for the three weeks he was willing to
spend with me at the University of Sussex, UK, where I designed the regional CGE
models used in this study.

The collaboration with the Agriculture Promotion Centre (CEPAGRI) and the
Department of Policy Analysis, Economics Directorate of the Ministry of Agriculture of
Mozambique and its Provincial branches of Tete and Sofala was also very important for
the completion of this work. I thank all the staff at the central and provincial levels, ﬁeld
supervisors, enumerators and support staff, whose contribution made this ambitious effort
possible. Particular recognition goes to my ﬁeld supervisors Arlindo Miguel, Julieta
Zandamela and Natércia de Sousa. I thank the cotton companies, C.N.A. and
DUNAVANT Mozambique, and the tobacco companies, MLT and DIMON
Mozambique, for the help provided in the organization and implementation of the ﬁeld
work at the district level, and by providing information crucial for the design and
implementation of this work.

This work is inevitably associated with the success of the Food Security III
Project, funded by USAID, that covered my Assistantship in the Department of
Agricultural Economics and a great deal of the ﬁeld work in Mozambique. In Spite of
not being formally part of the committee, Dr. Michael Weber, the Project Co-Director,
has always been a great source of wise advice throughout my Ph.D. Program. Additional
ﬁnancial support was obtained through a project funded by the AERC (African Economic
Research Consortium) that supported my staying at the University of Copenhagen,

Denmark, and University of Sussex, UK. In addition to funds from the Food Security

vi

Project, the second part of the ﬁeld work was generously funded by the Italian
Cooperation in Mozambique, through a funding facility obtained via CEPAGRI.
Additional funding on the initial phase was provided by the Ministry of Agriculture.

In the Department of Agricultural Economics, I would like to thank Dr. Eric
Crawford, the Graduate Program Advisor, for his advice throughout the program. A
special recognition also goes to Dr. John Staatz, whose class ‘Professional Practice in
Agricultural Economics’, a.k.a. ‘Boot Camp for Agricultural Economists’, was crucial in
increasing awareness and in improving my understanding of the practical challenges of
our profession and how to best go about them under different environments and
audiences; a golden key that I am pleased to keep!

The environment among graduate students at MSU was excellent. Without all the
friendship and fruitful interaction with my fellow graduate students, staff and professors,
it would have been hard to complete this journey successfully. My warm thanks to all!

I would also like to thank the Mozambican community of Graduate Students
across the US, particularly those at Purdue University, Louisiana State University,
Fresno State University, and especially Anabela Mabota at Ohio State University.
Together, we managed to make our stay enjoyable through the many visiting trips we
made to each other across the different States. I also thank our good friend Julia
Nhampule (who visited us in a critical moment during the program) for all the support.

Last but not least, I thank my wife Marcia and my son Ruy Allan for their love
and patience throughout the process. To my parents, Afonso and Maria Luisa, my
brothers and sisters, and my maternal grandmother who taught me the ABC, a big

Thanks!

vii

TABLE OF CONTENTS

LIST OF TABLES ............................................................. . .............................................. XI
LIST OF FIGURES ....................................................................................................... XIV
LIST OF APPENDIX TABLES .................................................................................. XVIII
LIST OF ACRONYMS ................................................................................................. XIX
CHAPTER 1

INTRODUCTION .............................................................................................................. 1
1.1. Background ............................................................................................................. 1
1.2. Research Objectives and Questions ........................................................................ 4
1.3. Organization of the Dissertation ............................................................................. 5
CHAPTER 2

FIELD RESEARCH METHODOLOGY ........................................................................... 7
2.1. Introduction ............................................................................................................. 7
2.2. Survey Sampling Approach .................................................................................... 7
2.3. Contents of the Survey Instruments ........................................................................ 9
2.4. Construction and Balancing of the Regional SAM ............................................... 10
CHAPTER 3

DISSERTATION ESSAY ONE: Interlinked Transactions in Cash Cropping Economies:
Rationale for Persistence, and the Determinants of Farmer Participation and Performance

in the Zambezi Valley of Mozambique ............................................................................ 13
3.1. Introduction ........................................................................................................... 13
3.2. An Overview of the Cotton and Tobacco Sub-Sectors ......................................... 17
3.2.1. The Tobacco Sector .......................................................................................... 17
3.2.2. The Cotton Sector ............................................................................................. 20

viii

3.3. The Organization of Production and Trade in Cotton and Tobacco Sectors ........ 24

3.3.1. Conceptual Framework ..................................................................................... 24
3.3.2. Contract Faming: Nature and Determining Factors .......................................... 29
3.4. Farmer Selection/Performance and Effects of Participation ................................. 34
3.4.1. Study Area Sample and Comparison of Means ................................................ 35
3.4.2. Conceptual Framework ..................................................................................... 42
3.4.3. Model Estimation and Discussion of Results ................................................... 49
3.5. Summary of Policy Implications .......................................................................... 69

CHAPTER 4

DISSERTATION ESSAY TWO: Income Poverty Effects of Expansion and Policies in

Cash Cropping Economies in Rural Mozambique: An Economy-Wide Approach ......... 74
4.1. Introduction ........................................................................................................... 74
4.2. Relevance for Pro-Poor Growth Policy ................................................................ 76
4.3. Data and Analytical Methods ................................................................................ 77
4.3.1. The Regional Social Accounting Matrix for the Zambezi Valley .................... 78
4.3.2. The Zambezi Valley Regional Computable General Equilibrium Model ........ 90
4.3.3. Interlinked Transactions and Expansion Paths ............................................... 111
4.3.4. Representative Households and Poverty Analysis in a CGE Framework ...... 113
4.4. Representative Characteristics of the Regional Economy .................................. 120
4.4.1. Introduction ..................................................................................................... 120
4.4.2. Demographics and Economic Structure .......................................................... 120
4.4.3. Structure of Production, Use of Factors and Intermediate Inputs ................... 122
4.4.4. Factor Incomes to Institutions and Structure of Household Income ............... 127
4.4.5. Structure of Household Expenditure ............................................................... 130
4.4.6. Domestic Supply and Demand, and Foreign Trade ........................................ 132
4.4.7. Poverty and Inequality Proﬁle ........................................................................ 136
4.5. Policy Simulations with the Regional CGE Model ............................................ 144
4.5.1. Introduction ..................................................................................................... 144
4.5.2. Policy Simulations in Tobacco Concession Areas .......................................... 152
4.5.3. Policy Simulations in Cotton Concession Areas ............................................ 162

ix

4.6. Summary of Policy “Implications ........................................................................ 173

CHAPTER 5
SUMMARY OF FINDINGS AND POLICY IMPLICATIONS .................................... 178
5.1. Introduction ......................................................................................................... 178
5.2. Summary of Findings .......................................................................................... 180
5.2.1. Organization of Production and Trade in Cash Cropping .............................. 180
5.2.2. Econometric Analysis: Farmer Selection/Performance and Effects of
Participation .................................................................................................... l 81
5.2.3. Economy-wide Analysis: Income Poverty Effects of Expansion
and Shocks ...................................................................................................... 186
5.3. Policy Implications and Future Research ........................................................... 189
APPENDIX A ................................................................................................................. 195
APPENDIX B ................................................................................................................. 200
REFERENCES ............................................................................................................... 213

Table 2.1.

Table 3.1.

Table 3.2.

Table 3.3.

Table 3.4.

Table 3.5.

Table 3.6.

Table 3.7.

Table 3.8.

Table 3.9.

Table 3.10.

Table 3.11.

Table 3.12.

Table 3.13.

LIST OF TABLES

Zambezi Valley Study: Survey Sample and Sample Attrition Rates ......... 8
Key Statistics of the Tobacco Sector in Mozambique, 2003-2004 .......... 18
Key Statistics of the Cotton Sector in Mozambique, 2003-2004 ............ 21
Transaction Cost Factors and Institutional Arrangements ........................ 28
Incidence of Selected Factors in Cotton and Tobacco Sectors ................. 32

Comparison of Mean Values for Selected Variables: Tobacco Growers
and Non-Growers ...................................................................................... 37

Comparison of Mean Values for Selected Variables: Cotton Growers and

Non-Growers ............................................................................................. 39

Mean Proﬁts. Net Crop and Total Income by Land Area Quartiles ......... 40

Mean Proﬁts, Net Crop and Total Income by Education Attainment

Level ......................................................................................... 41
Determinants of Proﬁts from Tobacco Production ................................... 53
Determinants of Proﬁts from Cotton Production ...................................... 58

Effects of Tobacco Contract Farming (CF) on Net Crop and Net Total
Household Income: Model with Land and Education Threshold Effects65

F-Tests of Joint Signiﬁcance of CF and Education and Land Thresholds:
Tobacco Areas .......................................................................................... 66

Effects of Cotton Contract Farming (CF) on Net Crop and Total
Household Income: Model with Land and Education Threshold Effects. 68

xi

Table 3.14.

Table 4.1.

Table 4.2.

Table 4.3.

Table 4.4.

Table 4.5.

Table 4.6.

Table 4.7.

Table 4.8.

Table 4.9.

Table 4.10.

Table 4.11.

Table 4.12.

Table 4.13.

Table 4.14.

F-Tests of Joint Signiﬁcance of CF and Education and Land
Thresholds: Cotton Areas ......................................................................... 69
Schematic Standard Regional Social Accounting Matrix ......................... 82
Regional SAM Accounts: Activities, Commodities and
Marketing Costs ........................................................................................ 85
Regional SAM Accounts: Factors, Institutions and
Savings-Investment ................................................................................... 86
Alternative Factor Market Closures ........................................................... 95
Tradability of SAM Commodities in the Zambezi Valley ...................... 106
Macro System Closures in the ZVR-CGE Model .................................. 109
Elasticity Sets and Deﬁnitions ................................................................ 1 11
Population Data for the Zambezi Valley Region .................................... 121
Structure of the Regional GDP ............................................................... 122
Aggregate Factor and Intermediate Input Shares by Activity ................ 124
Structure of Household Expenditures - percent ...................................... 131
Measures of Inequality ............................................................................ 142
Effects on Household Income of Alternative Simulations:
Tobacco Areas ........................................................................................ 148
Effects on Household Income of Alternative Simulations:
Cotton Areas ........................................................................................... 149

xii

Table 4.15.

Table 4.16.

Table 4.17.

Table 4.18.

Table 4.19.

Table 4.20.

Table 4.21.

Table 4.22.

Table 4.23.

Effects on Household Income of Alternative Simulations, Assuming
Unemployment and Full Mobility of Factors ......................................... 151

Base Shares and Changes in the Level of Activity by Simulation, Tobacco

Areas ....................................................................................................... 155
Changes in Demand for Factors from Actiyities in Tobacco Areas ....... 156
Base Factor Use Shares by Activity, Tobacco Areas ............................. 157

Base Shares and Changes in the Factor Remunerations, by Simulation 159

Base Shares and Changes in the Level of Activity in Cotton Areas ....... 166

Changes in Demand for Factors from Activities in Cotton Areas .......... 167

Base Factor Use Shares by Activity, Cotton Areas ................................ 167

Base Shares and Changes in the Factor Remunerations, by Simulation 170

xiii

Figure 3.1.

Figure 3.2.

Figure 3.3.

Figure 3.4.

Figure 3.5.

Figure 3.6.

Figure 3.7.

Figure 3.8. -

Figure 3.9.

Figure 3.10.

Figure 3.11.

Figure 4.1.

Figure 4.2.

LIST OF FIGURES

Raw Tobacco Production in Mozambique, 1996/97-2003/04 .................. 17

MLT-Tete and Dimon-Tete in Total Area and Production,
2003/2004 Season ..................................................................................... 19

Seed Cotton Production in Mozambique, 1994/95 — 2003/04 ................. 20

Dunavant-Tete and C.N.A.-Northern Sofala in Total Area and Production,
2003/2004 Season ..................................................................................... 22

Organization of Production and Trade: A Conceptual Framework .......... 27

Cotton Area Cultivated and Total Area Owned by Quartiles of Grower
Total Area ................................................................................................. 41

Tobacco Area Cultivated and Total Area Owned by Quartiles of Grower
Total Area ................................................................................................. 42

Cumulative Distributions of Tobacco Farmer Yields, by Firm, Zambezi
Valley — Mozambique, 2003/2004 ........... ‘ ............................................... 5 5

Cumulative Distributions of Tobacco Farmer Proﬁts, by Firm, Zambezi
Valley — Mozambique, 2003/2004 ............................................................ 56

Cumulative Distributions of Cotton Farmer Yields, by F irm, Zambezi
Valley — Mozambique, 2003/2004 ............................................................ 60

Cumulative Distributions of Cotton Farmer Proﬁts, by Firm, Zambezi

valley — Mozambique, 2003/2004 ............................................................. 61
Relationships Among Endogenous Accounts .......................................... 88
Stylized Regional Model Structure .......................................................... 91

xiv

Figure 4.3.

Figure 4.4.

Figure 4.5.

Figure 4.6.

Figure 4.7.

Figure 4.8.

Figure 4.9.

Figure 4.10.

Figure 4.11.

Figure 4.12.

Figure 4.13.

Figure 4.14.

Figure 4.15.

Figure 4.16.

Figure 4.17.

Figure 4.18.

Structure of Production Technology ........................................................ 93

Elasticities of Substitution ........................................................................ 94
Full Employment of Factors — Perfectly Inelastic Supply ........................ 96
Unemployment of Factors - Perfectly Elastic Supply ............................. 97

Flows and Speciﬁcation of Aggregation of Marketed Commodities ..... 102

Distribution Functions of Income per capita, by Household Group —
Illustrative Example Region 1 ........................................................... 118

Distribution Functions of Income per capita, by Household Group —
Illustrative Example Region 2 ................................................................ 1 19

Sectoral Structure of Value Added, Production and Employment ......... 123

Shares of Value Added, Production and Employment, by Commodity 125

Factor Shares Across Sectors of Activity .............................................. 126
Factor Payments to Domestic and Foreign Households ......................... 128
Shares of Factors and Transfers in Household Income .......................... 129
Structure of Household Expenditure ...................................................... 131
Sectoral Structure of Exports and Imports ............................................. 134
Commodity Structure of Exports ............................................................ 134
Commodity Structure of Imports ............................................................ 135

XV

Figure 4.19.

Figure 4.20. -

Figure 4.21.

Figure 4.22.

Figure 4.23.

Figure 4.24.

Figure 4.25.

Figure 4.26.

Figure 4.27.

Figure 4.28.

Figure 4.29.

Figure 4.30.

Shares of Exports in Output and Imports in Domestic Demand,
by Sector ................................................................................................. 136

Density Curves of Income per capita, by Concession Area, Zambezi
Valley - Mozambique, 2003/2004 .......................................................... 138

Distribution Curves of Income per capita, by Concession Area, Zambezi
Valley — Mozambique, 2003/2004 .......................................................... 139

Distribution Curves of Income per capita, by Household Group — Tobacco

Concession Areas .................................................................................... 140

Distribution Curves of Income per capita, by Household Group — Cotton
Concession Areas .................................................................................... 141

Lorenz Curves for Household Income per capita, by Household Group,
Zambezi Valley — Mozambique, 2003/2004 ........................................... 143

Comparison of Poverty Effects of Simulations, Tobacco Areas: Expansion,
Productivity and Export Price ................................................................. 152

Comparison of Poverty Effects of Simulations, Tobacco Growers:
Expansion, Productivity and Export Price .............................................. 153

Comparison of Poverty Effects of Simulations, Tobacco Non-growers:
Expansion, Productivity and Export Price .............................................. 154

Comparison of Poverty Effects of Simulations, Tobacco Areas: Expansion,
Import Price of Inputs and Export Tax ................................................... 160

Comparison of Poverty Effects of Simulations, Tobacco Growers:
Expansion, Import Price of Inputs and Export Tax ................................ 161

Comparison of Poverty Effects of Simulations, Tobacco Non-growers:
Expansion, Import Price of Inputs and Export Tax ................................ 162

xvi

Figure 4.31.

Figure 4.32.

Figure 4.33.

Figure 4.34.

Figure 4.35.

Figure 4.36.

Comparison of Poverty Effects of Simulations, Cotton Areas: Expansion,
Productivity and Export Price ................................................................. 163

Comparison of Poverty Effects of Simulations, Cotton Growers:
Expansion, Productivity and Export Price .............................................. 164

Comparison of Poverty Effects of Simulations, Cotton Non-growers:
Expansion, Productivity and Export Price .............................................. 165

Comparison of Poverty Effects of Simulations, Cotton Areas: Expansion,
Import Price of Inputs and Export Tax ................................................... 171

Comparison of Poverty Effects of Simulations, Cotton Growers:

Expansion, Import Price of Inputs and Export Tax ................................ 172

Comparison of Poverty Effects of Simulations, Cotton Non-growers:
Expansion, Import Price of Inputs and Export Tax ................................ 173

xvii

Table A].
Table B. 1.
Table B.2.

Table B.3.
Table B.4.

Table 3.5.
Table 3.6.
Table B.7.
Table 8.8.

Table B.9.

Table 8.10.
Table 8.11.
Table B.12.

Table 8.13.

LIST OF APPENDIX TABLES

Endogenous and Exogenous Sub-Matrices... ............................................ 198
Determinants of MSE Income, Two Stage Model in Tobacco Areas ....... 200
Determinants of Wage Labor Income, Two Stage Model

— Tobacco Areas ........................................................................................ 201
Determinants of MSE Income, Two Stage Model in Cotton Areas .......... 202
Determinants of Wage Labor Income, Two Stage Model

- Cotton Areas ........................................................................................... 203
Regional MACROSAM Structure (%) ..................................................... 204
Selected Characteristics of Zambezi Valley Smallholder Households ..... 205
Factor Shares Across Sectors of Activity .................................................. 206
Factor Shares within Sector of Activity: Agricultural, Livestock and
Fishing Activities ...................................................................................... 207
Factor Shares within Sector of Activity: Agricultural, Livestock and
Fishing Activities ...................................................................................... 208
Factor Income Distribution across Household Types ............................... 209
Shares of Factors and Transfers in Household income (%) ...................... 210

The Structure of Household Income in the Zambezi Valley
Smallholder Economy ............................................................................... 21 1
Economic Structure in the Base: Export and Import Shares ..................... 212

xviii

CDF

CES

CET

CF

CGE
CNA

CPI

DAD
DINA

F MC
FACTFE
FMOBFE
FMOBUE
GAMS
GDP
GOM
GTAP
IAF

IF PRI
IMR

INE

J F S

LEO

LES

MA

MLT
MPF
MSU
PRSP

PU

RH

RoW
SAM

S-I

TIA
ZVR-CGE
ZVR-SAM

LIST OF ACRONYMS

Cumulative Distribution Function

Constant Elasticity of Substitution

Constant Elasticity of Transformation

Contract Farming

Computable General Equilibrium

Companhia Nacional Algodoeira

Consumer Price Index

Distributive Analysis/Analyse Distributive Software
Direccao Nacional de Agricultura

Factor Market Closures

Factor is Activity Speciﬁc and Fully Employed
Factor is Mobile and Fully Employed

Factor is Mobile and Unemployed

General Algebraic Modeling System Software
Gross Domestic Product

Government of Mozambique

Global Trade Analysis Project

Inquérito aos Agregados F amiliares/Household Expenditure Survey
International Food policy Research Institute

Inverse Mills Ratio

Instituto Nacional de Estatistica

Joao Ferreira dos Santos

Leontief Technology

Linear Expenditure System

Ministério da Agricultura/Ministry of Agriculture
Mozambique Leaf Tobacco

Ministry ‘of Planning and Finance

Michigan State University .

Poverty Reduction Strategy Paper

Purdue University

Representative Household Approach

Rest of the World

Social Accounting Matrix

Savings-Investment

Trabalho de Inquérito Agricola/National Agricultural Survey
Zambezi Valley Region Computable General Equilibrium Model
Zambezi Valley Region Social Accounting Matrix

xix

CHAPTER 1
INTRODUCTION

1.1. Background

In spite of a signiﬁcant drop in recent years, more than half of Mozambique’s
population still lives under the poverty line. Absolute poverty is more prevalent in rural
areas where the great majority of the country’s population lives and works. Poverty in
those areas, using a consumption metrics, reaches about 55% of the population. Urban
poverty levels are also alarming; about 52% using the same metrics (MPF/IFPRI/PU,
2004).1 An analysis of changes in the poverty head count measure for the Zambezi

Valley region, the focus area of this study, indicates that the proportion of poor people
I fell more rapidly there than it did in the country as a whole between 1996/7 and 2002/3;
averaging a 28% drop across the provinces whose districts are included in the study
region. This contrasts with a 15% drop nationwide. A great deal of this recovery can
possibly be attributed to the expansion of cash cropping in the region in recent years,
through direct effects and indirectly from increased employment and consumption of
farm and non-farm goods and services, some of which are produced locally.

Research throughout the developing world has Shown a potentially strong
relationship between agro-industrial investments, growth in smallholder agriculture, and
poverty reduction (Jaffee and Morton, 1995; Dorward et al. 1998; Delgado, 1999). If
properly structured to relate to smallholder producers, those investments can play a very

important role in rural poverty reduction strategies in Mozambique (Tschirley, 2002). To

 

' Using the poverty headcount measure (Flexible Bundle Approach), national poverty incidence decreased
from 69% in 1996/7 to 54% in 2002/3. In rural areas, by the same measure, poverty reduced in 16
percentage points, against only 11 percentage points in urban areas nation-wide.

have a positive effect on rural poverty reduction in a sustainable manner, a necessary (but
not sufﬁcient) condition for any arrangement is that it must be proﬁtable for both the ﬁrm
involved and also for rural residents. Also, the issue of whom, within the income
distribution scale, grows speciﬁc crops and how proﬁtable they are, and who gets access
to employment opportunities, is essential to assess the direct impact on income
differentiation of the arrangements or technological options. Growth in rural incomes as a
result of indirect effects depends on the patterns of demand and the structure and
ﬂexibility of response of the farm and non-farm economy to changes in effective
demand.2

In Mozambique, cotton and tobacco are important cash crops grown by the
smallholder sector and demanded by a large-scale processing/trading sector with
important links to external markets. These are both very demanding crops in input and
factor use, particularly chemical inputs and labor; this is particularly important in
tobacco, which uses a wider range of productivity enhancing inputs and is more
demanding in its ﬁeld practice requirements. Given the current stage of development of
rural agricultural input and credit markets in the country, farmers have little access to
improved seed and chemical inputs in rural areas. These market failures result in a heavy
reliance on contract farming as the dominant mode of sub-sector organization.

Most schemes take the form of forward resource/management contracts.3 The

 

2 For a discussion on supply chain institutional arrangements and poverty reduction in rural Mozambique
see Benﬁca er al, 2002.

3 These differ from the simple sale/purchase contracts because they include stipulations regarding the
transfer and use of speciﬁc resources and/or managerial functions (Jaffee and Morton, 1995). Forward
resource management contracts partially internalize product and factor transactions, and are sometimes
referred to as interlinked contracts or interlocked markets (Minot, 1986; Glover and Kusterer, 1990;
Dorward et al. 1998).

standard contract consists in the ﬁrms supplying seeds, other inputs (including
chemicals), and technical assistance to farmers on credit for use on Speciﬁc crops.
Farmers agree to utilize the inputs as instructed, and to sell all their production to the
ﬁrms at harvest at agreed-upon prices. The costs initially supported by the ﬁrms,
together with any interest charge, are deducted at the time of the harvest. In the Zambezi
Valley of Mozambique, government has granted the cotton and tobacco companies
monopsony rights, i.e., farmers are not permitted to sell outside the contracting scheme
they are assigned to. These schemes are normally implemented on land for which
individual farmers or farming communities have use rights, but there are cases in other
parts of the country where the ﬁrms use designated areas or blocks within their own land
concessions for that purpose (Strasberg, 1997).

Results from recent studies in the Zambezi Valley (Walker et al., 2004; Benﬁca er
al. , 2005) indicate that growers of cash crops, particularly cotton and tobacco, tend to
have crop and total incomes somewhat higher than those bf non-growers.4 Those studies
have also indicated that a number of farmers engaged in the contracting schemes have
losses in their operations. On the other hand, the extent of indirect effects through labor
markets and other linkages, which can potentially foster broader income distribution and
impact poverty reduction across different household groups, have not been fully
documented.

This study is motivated by three researchable hypotheses. First, resource and
technology endowments are important for access to and performance in contract farming

in cotton and tobacco areas. Second, due to farmer heterogeneity, the effects of

 

‘ Differences in total incomes are much less accentuated due to relatively higher non-agricultural incomes
earned by nonocash crop growers (Tschirley and Weber, 1994; Benﬁca et al., 2005). This may imply more
equitable effects of cash crops on a broader scale.

participation in contract farming for speciﬁc types of households can differ substantially
from the average effect across all households; understanding what type of farmer beneﬁts
from participation is crucial for designing effective policy for the sectors, and can also be
useful in advising companies on effective expansion paths. Finally, the emerging trends
in non-crop incomes and the incidence of labor market interactions and farm/non-farm
linkages, associated with a dynamic cross-border economy, seem to indicate that contract
farming can be inequality reducing, i.e., non-growers of those crops, in some areas, may
be indirectly affected by that growth and the policies and exogenous shocks in cotton and

tobacco sectors in the Zambezi Valley.

1.2. Research Objectives and Questions

The general objective of this study is to help guide government and donors in
allocating development resources aimed at inducing private sector investments to take
place in ways. expected to yield a high proﬁt potential while generating signiﬁcant
income growth and poverty reduction effects in rural areas of the Zambezi Valley of
Mozambique.

The study has three speciﬁc objectives and associated research questions:

1. To understand the rationale for persistence and the determinants of farmer
participation and performance in cotton and tobacco contract farming schemes in

the Zambezi Valley of Mozambique. The associated research questions are:

a. Why is contract farming the dominant and persistent institutional
arrangement in the Zambezi Valley cash cropping economies?

b. What are the determinants of farmer participation in contract farming
schemes in cotton and tobacco sectors?

c. What determines participant farmer proﬁtability in those schemes?

2. To assess the effects of participation in contract farming on agricultural and total
household incomes, accounting for threshold effects of education and land

holdings, and controlling for selectivity bias. We aim at answering the questions:

a. To what extent does participation in contract farming explain differences
in farmer crop and total household incomes?

b. What types of farmers beneﬁt from contract farming?

3. To understand the mechanisms and assess the economy-wide income and poverty
effects of expansion, policies and Shocks to cash cropping sectors on household

groups. Speciﬁc questions include:

a. What kinds of policies have the greatest impact on poverty reduction?

b. What is the best combination of policies to counteract negative shocks?

1.3. Organization of the Dissertation

This dissertation is organized in ﬁve chapters, including this introduction. Chapter
Two details the methodology used in the data collection process, and discusses the
construction and balancing of the Social Accounting Matrix used in the economy-wide
analysis.

In Chapter Three, we present the ﬁrst Dissertation Essay that addresses the ﬁrst
two research Objectives. In the ﬁrst part of the Essay, we present an overview of the
cotton and tobacco sectors, followed by an analysis of the organization of production and
trade, to review the rationale for the dominance and persistence of contract farming in
those sectors. Then, two-stage econometric procedures (a sample selection and a
treatment effects model with land and educational attainrrient thresholds) are used to

assess the determinants of farmer participation and performance in contract farming, and

the effects of participation on crop and total household incomes of farming households in
cash cropping areas. Finally, the chapter addresses speciﬁc policy implications for each
area.

Chapter Four introduces the second Dissertation Essay that addresses the third
research objective of the study. First, it presents the detailed structure of the regional
SAM and the CGE model, commodity balances, factor market and macro system
closures, as well as considerations on model calibration. Second, it uses the regional
SAM to derive a set of representative characteristics of the economy. Finally, the
economy-wide framework is used to assess the income and poverty effects of expansion
and alternative policies and Shocks in the cash cropping sectors on growers and non-
growers in those economies, using cumulative distribution functions of household income
per capita.

Chapter Five presents a summary of conclusions, economic policy implications

and associated recommendations for further research.

CHAPTER 2
FIELD RESEARCH METHODOLOGY
2.1. Introduction

This chapter presents the methodology used for data collection in this study. To
get sufﬁciently reliable data for the analysis, a multi-visit survey was undertaken in the
study region. Two visits, six months apart, allowed for more precision in the data
collected on the levels of input use and the variation in factor use, particularly the
seasonality in labor demand and household decisions with respect to the use of family or
wage labor and its allocation across competing activities. The schedule for the ﬁeld data
collection was driven by the crop calendar of the two cash crops of interest. The data
collected were used for both the econometric analysis in Chapter Three and for the
construction of the Social Accounting Matrix (SAM) used as the database in the
economy-wide analysis in Chapter Four. The following are details on the sampling
strategy and coverage, the contents of the survey instrument, and considerations

regarding the construction and balancing of the SAM.

2.2. Survey Sampling Approach

The survey followed a stratiﬁed random sampling procedure. It covered
concession areas for four ﬁrms operating contract farming schemes in the Zambezi
Valley of Mozambique. Two of the ﬁrms, Mozambique Leaf Tobacco (MLT) and
DIMON Mozambique, operate tobacco growing schemes in Tete Province. The other two
are cotton companies, Dunavant Mozambique operating in Tete Province and C.N.A.

operating in Northern Sofala Province.

The survey targeted a total of 300 smallholder farmers interviewed in two rounds,
each round recalling six months of the 2003/2004 agricultural season. In tobacco areas,
the sample targeted 180 farmers among growers (130) and non-growers (50). One
hundred of those farmers were drawn from the Mozambique Leaf Tobacco Area (70
growers and 30 non-growers), and 80 were from the DIMON Mozambique area (60
growers and 20 non-growers). In cotton areas 120 farmers were targeted, 90 growers and
30 non-growers. Sixty of those farmers were drawn from the Dunavant Mozambique
Area (42 growers and 18 non-growers), and the other 60 were from the C.N.A. area (48

growers and 12 non—growers).

Table 2.1. Zambezi Valley Study: Survey Sample and Sample Attrition Rates

 

 

 

 

 

 

 

First Round of Survey Second Round of Survey Sample
Growing Areas — Number of Farmers — — Number of Farmers —— MES?“
And Firms
Growers Non- Total Growers Non- Total (%)
Growers Growers
Tobacco Areas
MLT 70 30 100 64 26 90 10.0
DIMON 60 20 80 53 16 69 13 .8
All Area 130 50 180 117' 42 159 11.7
Cotton Areas
Dunavant 42 18 60 4| 1 8 59 1 .7
C.N.A. 48 12 60 46 12 58 3.3
All Area 90 30 120 87 30 117 2.5
All Sample 220 80 300 204 72 276 8.0

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

The households were ﬁrst interviewed in March 2004 to collect data on ﬁeld and
economic activities for the period September 2003 — February 2004, which covers pre-

harvesting activities for most crops. The second round was implemented in September

2004 to collect data for the remainder of the period that included harvesting and post-
harvesting activities. Because tobacco harvesting started around February/March and
continued for several months, information on sales and data on labor use for harvesting
and marketing activities for this crop were collected in the second round.

Table 2.1 shows a sample attrition rate of 8.0%, which reduced our sample to 276
observations.5 To ensure appropriate statistical treatment of the data collected when
generating the results, we developed sampling weights derived from the sample selection
probabilities resulting from the sampling strategy and available population data. Those
weights are used in the generation of all results in the statistical and econometric analysis
and for generating representative aggregates at the regional level when constructing the

SAMs.

2.3. Contents of the Survey Instruments

The objective of the survey was to get household level data on the use of
intermediary inputs and factors, and production, sales and earnings information for the
major agricultural and non-agricultural activities undertaken in the selected region.
These include the relevant cash crops, cotton and tobacco, other agricultural production
sectors and selected non-farm activities as well as details on other sources of income,
asset ownership and investments. Multi-level survey questionnaire design techniques
were applied to allow for appropriate treatment of the data collected in the various parts

of the survey. Excepting for minor differences, the survey instruments used in both

 

5 In panel data collection, where the same households are visited multiple times, the sample attrition rate
refers to the proportion of households that are not re-interviewed for a variety of reasons.

rounds were very similar, but in each case applied for a speciﬁc recall period. The

following is a list of topics included in the survey instrument:

Smallholder household demographics;

Land use and cropping patterns;

Use and sources of farm non-labor inputs;

Labor allocation on farm and non-farm activities;
Production and marketing of crops;

Ownership of production and marketing assets;
Ownership and marketing of livestock;

Income diversiﬁcation: Micro and small enterprises and wage labor;
Remittances from (and to) rural smallholder households;
Pensions and other transfers;

Uses of cash revenues: Assets and business startups;

Smoking habits in rural smallholder households.

Electronic copies of the questionnaires and data sets are available on the Michigan

State University Food Security Group website.6

2.4. Construction and Balancing of the Regional SAM

The Regional SAM is the database used in the economy-wide analysis.

Assembling a SAM is a complex process that requires a great deal of perseverance and a

strong dose of ingenuity (Dervis et al., 1982; Thorbecke, 1998; Sadoulet and de Janvry,

1995; Taylor and Adelman, 1996). The process of building the Zambezi Valley Regional

SAM (ZVR-SAM) included several steps. First, we designed a speciﬁc SAM

framework, that consisted in the identiﬁcation of the relevant accounts, and the

appropriate level of dis-aggregation reﬂecting the structure of the local economy and the

 

6 httn://www.aec.msu.edu/st/mozambique/survev/index.htm

10

objectives and issues identiﬁed for the research. Second, based on the anatomy deﬁned
for the various accounts, we explored in detail the existing sources of data, identiﬁed the
gaps, and the needs for additional data collection.

Given the speciﬁc nature of the research and the limited availability of data at the
sub-national level, a great deal of the data used to construct the ZVR-SAM, particularly
on the production and income Sides, had to be collected through the sample survey, as
described in the previous section. Expansion procedures using population census data
(INE, 2000) were necessary to make the data representative of the study area.

On the consumption side, we derived household expenditure shares using the
National Expenditure Survey (IAF 2002-03) for the enumeration areas corresponding to
our study region. The data were used to estimate household home consumption and
expenditures across commodities in the SAM for the different household groups. To
accommodate those data in the SAM, a matching procedure was adopted. It consisted in
ranking the four household groups in the SAM by total income, then attributing to each
the average expenditure pattern of the relevant quartile group from the IAF data for the
Zambezi Valley: the poorest group in the SAM received the expenditure pattern of the
bottom quartile from IAF, through to the richest group in the SAM, which received the
expenditure pattern of the top quartile in IAF. The analysis of production, marketing, and
domestic consumption data allowed for the derivation of variables such as crop exports
and household transfers to the rest of the world.

The end result of this work was a consolidated SAM representative of the
Zambezi Valley Region. Additionally, for analytical purposes, we disaggregated the

master SAM into two Sector/Concession speciﬁc SAMS.

ll

Despite the careful data collection, processing and cleaning, and a very interactive
process in assembling the SAMs, inconsistencies remained, arising from measurement
errors, incompatible data sources, and lack of data. To impose consistency, we used the
Cross-Entropy Method (Robinson, et al., 1998). This method is based in information
theory and incorporates errors in variables, inequality constraints and prior knowledge
about any part of the SAM including, but not restricted to, rows and column sums. All the
necessary adjustments resulting from the procedure were within the generally acceptable

bounds of less than 5%.

12

CHAPTER 3
DISSERTATION ESSAY ONE

Interlinked Transactions in Cash Cropping Economies: Rationale for Persistence,
and the Determinants of Farmer Participation and Performance in the Zambezi
Valley of Mozambique

3.1. Introduction

Contract farming is the most pervasive form of market organization in cash crop
production in Mozambique. All cotton and tobacco production by smallholder growers in
the Zambezi Valley region originates from contract farming arrangements promoted by
agro-industrial ﬁrms. In total, there are two out-grower tobacco companies and three
ﬁrms devoted to the management'of cotton contract farming schemes. Those ﬁrms are
assigned speciﬁc geographical “concession” areas, where they provide inputs and
extension assistance to small farmers on credit, and are granted monopsony rights that
entitle them to purchase all the output at predetermined prices.

Contract coordination, through contract farming interlinked transactions, present
several potential advantages. First, it helps to cope with market failure by reducing
uncertainty for farmers regarding access to inputs, services and output markets (Glover,
1984; Goldsmith, 1985; Minot, 1986) and by assuring access to sufﬁcient raw material of
acceptable quality for processors. Second, it may signiﬁcantly raise income of growers
and enhance rural development by serving as a source of information for new production
technologies. Finally, it may trigger multiplier effects through employment linkages, and
infra-structure and marketing development in the local economy (Warning and Key,

2002).

I3

There are also potential limitations and negative impacts of contract farming.
First, given its monopsonic nature, these arrangements may result in asymmetric
bargaining where one buyer largely determines the prevailing price and contract
conditions. After farmers have invested in speciﬁc assets, or altered their cropping
patterns and become more dependent on their contract crOpS, they may lose bargaining
power and be more likely to accept less favorable or exploitative contract terms (Little
and Watts, 1994). Second, the cost of enforcing contract provisions can be very high for
both parties due to opportunistic behavior by participants and weaknesses in the legal
system in rural areas. Finally, contract farming may result in barriers to entry for farmers
when processors limit suppliers to those capable of meeting volume and quality standard
requirements, typically the already better off farmers (Benﬁca, Tschirley and Sambo,
2002).7

In addition to the quality and intensity of the assistance and the prevailing prices
and overall world market conditions, the performance of contract farming schemes and
their broader impacts on rural development depends on the types of growers that get
contracted. It is clear that if ﬁrms contract primarily with wealthier growers, the poorer
members of the community will fail to beneﬁt directly ﬁ'om the contract arrangements
(Warning and Key, 2002). Nevertheless, the extent to which such approaches will
exacerbate existing patterns of economic stratiﬁcation, as argued by Key and Runsten
(1999), cannot be conclusive without further investigation. The net effect depends upon

the extent of economic linkages. In the Zambezi Valley case, at least three effects are

 

7 In addition to those limitations, widespread contractual coordination may raise price volatility in the
remaining spot market transactions, due to the thinness of those markets and the lack of transparency across
many contracts within the same sub-sector. This will reduce or distort the information supplied by those
spot market prices.

14

especially important: (i) spillover effects on food crop yields through increased fertilizer
use, and input market development; (ii) the effects of increased labor demand and re-
spending of wage earnings; and (iii) multiplier effects from re-spending of cash crop
earnings. The presence of these important spillovers from growers to non-growers makes
the assessment an empirical question.

The objective of this paper is as follows. First, we develop a conceptual
framework following Williamson (1991) and link it to an institutional analysis to identify
the factors determining the dominance of contract farming in cotton and tobacco value
chainsin the Zambezi Valley region. Second, taking into account the selective nature of
participation in those schemes and the stratiﬁed random nature of our sample, we develop
two versions of sample selection models (Heckman, 1979; Greene, 2003) to address the
following issues: (i) the determinants of farmer selection into the contract farming
schemes; (ii) once selected into the schemes, the determinants of participants’
performance; and (iii) assess whether overall agricultural and total household income of
participants is statistically higher than that of non-participants after controlling for
demographic, factor and asset endowments, structural factors, and sample selection bias.
A key contribution of this paper is its investigation of threshold effects of education and
land holdings; rather than focusing on the average effect of participation, we ask what
type of farmer beneﬁts from participation; an answer to this question is crucial for policy
design and to shed light on the identiﬁcation of effective expansion paths to companies.

While sample selection and treatment effects models are common in many areas,
applications to the performance and impact of contract farming in developing countries

are rare. Warning and Key (2002) used a treatment effects model to analyze the impact

15

of the Arachide de Bouche Confectionary Peanut Program in Senegal. They found that
program participants and non-participants were indistinguishable by wealth measures and
that participating farmers increased their gross agricultural income substantially. One
limitation of their study was the extremely small sample size; only 26 observations. Our
study is an empirical contribution to the literature in three ways. First, it uses a much
larger sample size and controls for a larger set of variables than previous studies. Second,
we use interaction terms to assess how participation effects vary across thresholds of land
holdings and education. Finally, we extend the analysis to include a standard sample
selection model and evaluate the determinants of farmer selection and performance
within the contract farming scheme itself.

This essay is organized as follows. Section 3.2 gives an overview of the tobacco
and cotton sub-sectors in Mozambique, including some measures of its recent
performance, and puts the study region into perspective. Section 3.3 develops a
conceptual framework for the analysis of the organization of production and trade in the
value chains, and identiﬁes the nature of contract farming and the factors leading to its
dominance in the Zambezi Valley region. Section 3.4 reviews the study area sampling
coverage and undertakes a comparison of means between scheme participants and non-
participants for selected household level variables. Then, it outlines the sample selection
problem and correction, and the speciﬁcation of the cash crop income determinants
model and the treatment effects model with sample selection correction and land holdings
and education threshold effects. Finally, it presents the estimation and the discussion of

results for both models. Section 3.5 closes with the discussion of policy implications.

16

3.2. An Overview of the Cotton and Tobacco Sub-sectors

This section presents an overview of the tobacco and cotton sectors in
Mozambique. It starts by looking at national production trends over the past decade.
Then, it characterizes the 2003/2004 agricultural season by province and by ﬁrm, and

puts the study region into perspective.

3.2.1. The Tobacco Sector

Tobacco production in Mozambique has grown very rapidly over the past decade.
From 1,500 tons in the 1996/7 agricultural season, national production of raw tobacco has
increased every year to reach over 54,000 tons in 2003/4; see Figure 3.1. Over the same
period the estimated number of tobacco growing households has increased from 6,000 to

more than 100,000.

Figure 3.1
Raw Tobacco Production in Mozambique

1996/97-2003/04

Metric Tons

 

 

0 I I I I I I I j

96/97 97/987 '98/99 99/00 00/01 01/02 02/03 03/04
Agricultural Seasons

Source: DINA-MA

17

There are currently ﬁve major F irms/Partnerships operating in the country
promoting both smallholder contract farming schemes and larger scale commercial
operations.8 The positive impact of this rapid expansion in the tobacco sector on rural
smallholder household incomes and welfare has been dramatic (Walker et. al, 2004;
Donovan, 2004; Boughton et al., 2004; Benﬁca et al., 2004, 2005). Table 3.1 presents

key tobacco sector statistics for the agricultural season 2003/2004.

Table 3.1. Key Statistics of the Tobacco Sector in Mozambique, 2003-2004

 

 

 

 

 

 

 

 

Area Planted Production
Province/Firms Yield
Area (ha) % Volume (tons) % (tons/ha)

All Mozambique 62,315 100.0 54,408 100.0 0.87
Total by Province
Niassa 8,977 14.4 7.692 14.1 0.86
Cabo Delgado . 82 0.1 82 0.2 1.00
Nampula 5,985 9.6 3,625 6.7 0.61
Zambezia 3,991 6.4 2,391 4.4 0.60
Tete 32,381 52 0 27,032 49 7 O 84

MLT 23,849 38.3 20,000 36 8 0 84

DIMON 8,532 13.7 7,032 12 9 0 82
Manica 10,359 16.6 13,214 24.3 1.28
Sofala 510 0.8 360 0.7 0.71
Gaza 30 0.0 12 0.0 0.40
Total by Firm

JFS 13,127 21.1 9,341 17.2 0.71

MLT 25,288 40.6 22,920 42.1 0.91

DIMON 12,594 20.2 10,950 20.1 0.87

STANCOM 11,306 18.1 1 1,197 20.6 0.99

 

Notes: The total number of growers in the country is estimated at over 100,000. From those, 44,783 work
in Tete Province alone (MLT: 34,038 and DIMON-Mozambique: 10,745). Due to the absence of precise
data for most of the ﬁrms. data on the number of producers is not detailed in the Table.

Source: DINA-MA, and Individual Firms.

 

a The Firms/Partnerships operating in the country are as follows: MLT — Mozambique Leaf Tobacco (Tete
and Manica); JFS - Joao Ferreira dos Santos (Manica, Nampula, Cabo Delgado, Niassa, and Gaza);
DIMON (Tete, Manica, and Sofala); Stancom/Mosagrius (N iassa); and Stancom/Sonil (Nampula).

18

Overall, there were 62,315 hectares of land planted with tobacco in eight
Provinces. That area includes commercial farming by large growers and smallholder
growers involved in contract farming schemes. Total production in that season reached
over 54,000 tons of raw tobacco, of which about three quarters are of the burley type. The
total number of growers is estimated at over 100,000.

About 52% of the total area planted nationally and 50% of the total production
was by smallholder growers in our study region (Tete Province). Those farmers were
engaged in contract farming schemes with Mozambique Leaf Tobacco (3 7% of the
national production by 34,038 farmers), and Dimon Mozambique (13% of the national

production by 10,745 farmers).9 See percentage distributions of land area and production

     

         

 

  

 

in Figure 3.2.
Figure 3.2
MLT-Tete and Dimon—Tete in Total Area and Production
2003/2004 Season

120 -
._ 100 7 \2
‘3 80 . \\
I—
c... 60 .
O . 7 \

o- nmmmn ﬂ§ :

 

Area Cultivated (In) Production (tons)

E MLT-Tete III Dimon-Tete Other Areas Total

Source: DINA-MA and tobacco companies

 

9 Note that if we account for the production those two ﬁrms get from their global operations in the country,
their national production share is much higher: MLT (42%) and DIMON-Mozambique (20%).

I9

3.2.2. The Cotton Sector

Cotton production in Mozambique has varied widely over the years. Current
production is dramatically higher than from the early- to mid-19805, but is only about
60% of the historical high, achieved more than 30 years ago. Key factors explaining the
decline are persistently low producer prices, the lowest in the region, and issues related to
the organization and performance of the contract farming arrangements between
ginning/exporting ﬁrms and smallholder farmers. The production of 88,000 tons achieved

in 2003/2004 still falls short of the ten-year high achieved in 1998/99 (Figure 3.3).

Figure 3.3

Seed Cotton Production in Mozambique
1994/95 - 2003/04

Metric Tons
,8

 

 

O T I I I I T I I T I
94/95 95/96 96/97 97/98 98/99 99/00 00/01 01/02 02/03 03/04

Agricultural Seasons

Source: 1AM — Instituto do Algodao de Mocambique

Historically, cotton production in Mozambique has been concentrated in the
northern part of the country, particularly in Nampula Province, but that pattern is

currently changing. Between 1990 and 2000, Nampula production accounted, on average,

20

for 52% of national production, against only 12% for all central provinces (Tete,

Zambézia, Sofala, and Manica). By 2003/04 the ﬁgures were 39% and 29%, respectively.

This shifting balance has been driven by continuing problems in Nampula, and by the

entrance of two new ﬁrms in the Center committed to increasing yields and overall

development of the supply chain (Tschirley et al. , 2005).

Table 3.2. Key Statistics of the Cotton Sector in Mozambique, 2003-2004

 

 

 

 

 

 

 

 

 

Growers Area Planted Production Yield
Province/F inns (tons/ha)
Growers Area (ha) % Volume (tons) %
All Mozambique 107,845 174,157 100.0 88,173 100.0 0.51
Provinces
Niassa - 14,863 8.53 7,817 8.87 0.53
Cabo Delgado 41,671 38,958 22.37 20,819 23.61 0.53
Nampula 24,759 75,606 43.41 34,144 38.72 0.45
Zambézia 9,918 13,957 8.01 3,940 4.47 0.28
Tete 7,430 7,361 4.23 4,256 4.83 0.58
DUNAVANT 4,022 2,257 1.30 1,037 1.18 0.46
AGRIMO 3,408 2,627 . 1.51 1,839 2.09 0.70
COTI‘CO - 2.477 1.42 1,380 1.57 0.56
Manica 1,685 7,531 4.29 4,067 4.61 0.54
Sofala (C.N.A.) 22,382 15,937 9.15 13,130 14.89 0.82
Total by Firm
SAN/JFS - 18,923 10.87 11,208 12.71 0.59
PLEXUS 32,691 31,312 17.98 17,485 19.83 0.56
SODAN 32,987 25,430 14.60 10,544 1 1.96 0.41
MOCOTEX 215 2,739 1.57 427 0.48 0.16
SANAM - 31,047 17.83 1 1,137 12.63 0.36
CANAM - 20,460 1 1.75 10,774 12.22 0.53
IAM/MEMBA 470 400 0.23 48 0.05 0.12
AGRIMO 13,326 8,434 4.84 4,770 5.41 0.57
SAAM - 5,500 3.16 607 0.69 0.1 1
DUNAVANT 4,022 2,257 1.30 1,037 1.18 0.46
COTTCO - 8,758 5.03 4,562 5.17 0.52
C.N.A. 24,067 17,131 9.84 14,015 15.89 0.82
Autonomous 67 1,766 1.01 1,559 1.77 0.88

 

Notes: DUNAVANT and AGRIMO merged recently.

Source: 1AM and Individual F irrns.

21

In terms of individual ﬁrms, in 2003/04 (Table 3.2), Plexus (in Cabo Delgado
province, north of Nampula) accounts for 18% of the area cultivated and 20% of total

production. CNA in the Center is second with about 16% of national production, in spite

of cultivating only 10% of national cotton area. '0

Unlike tobacco, cotton production in our study area does not represent the
dominant Share of national production. While C.N.A. began operations over a decade
ago, the other two companies in the area (Dunavant Mozambique and COTTCO/Algodao
do Zambeze) started only within the past three years. All together, the ﬁrms included in
the survey sample (Dunavant/Tete and C.N.A./Northern Sofala) account for about 18,000
hectares, or 11% of the national cotton area, and 14,000 tons, approximately 17% of the
national production in 2003/2004 (Figure 3.4). In the 2003/4 season C.N.A. worked with

approximately 22,000 growers, while DUNAVANT operated with about 4,000.

Figure 3.4

Dunavant-Tete and C.N.A.-Northern Sofala
in Total Area and Production
2003/2004 Season

% of Total

 

 

 

Area Cultivated (ha) Production (tons)
a Dunavant-Tete Ill CNA-N.Sofala Other Areas Total

Source: DINA-MA and cotton companies

 

'0 Yields by CNA farmers (0.82 tons/ha) are well above the national average of 0.51 tons/ha in 2003/04.

22

Overall, the performance of the cotton sub-sector in Mozambique has been far
from satisfactory (The World Bank, 2005; Tschirley et al., 2006). Farmer proﬁts remain
well below potential due to very poor yields and low producer prices. In fact,
Mozambique pays the lowest prices in the region; the 1998-2002 average producer prices
were $0.16 per kilogram, compared to $0.22 in Zambia and Tanzania and $0.25 in
Zimbabwe (Poulton et al., 2004). Yields in 2003/04 were 0.51 tonnes per hectare,
compared to 0.9 tonnes in Zimbabwe and over 1.0 ton in West Africa (Lemaitre et al.,
2001). While C.N.A. presents the best yield record in the country (over 0.8 tonnes/ha), it
pays farmers statutory minimum prices, well below those in other countries. The other
player in the sample, Dunavant Mozambique, has only recently started operations; its
yields are still relatively low, but it makes an effort to retain farmers by paying prices
well above the national average. Given. its successful track record in Zambia, it is
expected to play a key role in the development of the Mozambican cotton sector.

Further down the value chain, a large proportion of installed ginning capacity
remains unutilized, and ginning outtums and lint quality are very low. Since the late
19903, ginning outtums averages 35% in the country, compared to 38% in Zambia 40%
in Zimbabwe and 42% in West Africa (Lemaitre et al. , 2001; Oﬁco and Tschirley, 2003;
Horus, 2004; and World Bank, 2005).11 As a result of its low quality, the price paid to
Mozambique cotton in the world market is signiﬁcantly discounted relatively to the Index
A price. Therefore, improvements in ginning quality and yields can have positive effects
on the competitiveness of the sector by reducing lint production costs and improving

export prices (World Bank, 2005), increasing the prospects for better prices to farmers.

 

" More recently ginning outtums reached around 37%, with C.N.A. achieving as high as 41.5% in the
2001/02 season.

23

3.3. The Organization of Production and Trade in Cotton and Tobacco Sectors
3.3.1. Conceptual Framework

The approach used to analyze the organization of production and trade in this
section follows Williamson in the sense that institutions are explicitly endogenized,
particularly the process of institutional change and the choice and design of institutional
arrangements (Williamson, 1991). Institutions are central to economic development
because they affect production and transaction costs (North, 1990). In the presence of
transaction costs and information constraints, institutions inﬂuence the efﬁciency and
distribution of resources (Cook and Chaddad, 2000). This approach is particularly
relevant in developing countries, where high transaction costs, missing markets and
market failures]2 are the rule rather than the exception (Bardhan, 1989).

The likely effect of agro-industrial investment on smallholder welfare is to some
extent related to the nature of the institutional relationship (contractual form) between
farmers and agro-industrial ﬁrms. Williamson (1991) identiﬁes three broad types of
contractual forms: neo-classical spot markets, bilateral contracts, where autonomous
parties enter into contracts that extend beyond single transactions, and vertical integration
within a ﬁrm.

Transactions underlie each of these institutional arrangements. In this context,
transactions refer to the activities that allow or constrain transformation activities. A
transaction occurs when two or more parties enter into an arrangement in which rights

and obligations are exchanged (Staatz, 1988). All transactions come bundled with a mix

 

'2 Missing markets refer to a situation where there is no market to govern the allocation of resources or
goods and services. Market failure is a situation in which markets exist but do not allocate resources
efficiently. Market failure may be due to market power, extemalities, public good nature of the goods or
services, or the existence of incomplete or asymmetric information, and uncertainty.

24

of characteristics - the degree of asset specificity, the degree and type of uncertainty to
which the parties are subject to, and the complexity and frequency with which the
transactions occur (Williamson, 1991). For example, transactions which occur under
isolated spot markets for low value commodities involve relatively low levels of all these
characteristics. Transactions underlying a contract farming scheme have higher levels of
these characteristics.

The mix of transaction characteristics is inﬂuenced by a number of factors related
to production, marketing and processing characteristics, and to factors related to the
economic and political environment (Jaffee and Morton, 1995; Delgado, 1999; and
Benﬁca, Tschirley and Sambo, 2002). Competitive forces tend to promote the emergence
of forms of economic organization that minimize total costs of production and exchange
in the economic system (Staatz, 1988).

The analytical model used in this research is a simpliﬁed version, although
following the same logic, of models used in much of the applied work in transaction costs
economics (Klein, 1995; and Dorward, 2001). The efﬁcient form of organization for a
given economic relationship, i.e., the likelihood of observing a particular organizational
form, is a function of certain properties of the underlying transactions. Formally: Y = (D
[X], where, Y is a vector of alternative arrangements/organizational forms, more
speciﬁcally: Spot marketing, contract farming alternatives, and plantation agriculture;
and X is a vector of transaction characteristics that affect transaction costs, more
speciﬁcally: condition and degree of asset speciﬁcity, degree and type of uncertainty that
parties to the transaction are subject to, and complexity and frequency with which they

occur.

25

The probability of observing a more integrated organizational form depends
positively on the amount or value of the speciﬁc assets involved in the relevant
transactions, on the degree of uncertainty, the complexity of the transaction and its
frequency (Klein, 1995). In general, therefore, the greater the degree of asset speciﬁcity,
the less likely it is that spot markets will be relied upon. In that case, contractors will seek
to negotiate contracts that protect their investment in the face of external change. Figure
3.5 illustrates that low degrees of uncertainty, complexity and frequency may favor spot
markets and reduce the need for vertical coordination. The opposite, however, may lead
to the recognized need of building contractual relationships that acknowledge mutual
interest in contracting, facilitate information ﬂows and allow for a ﬂexible joint response
to changes in external circumstances; this includes a wide range of contractual
arrangements.” But such relationships require trust. Where trust cannot be established,
vertical integration may. be chosen instead (Dorward, Kydd and Poulton, 1998).

The level of the elements in X is inﬂuenced by a number of factors related to
production characteristics, marketing/processing characteristics and the economic and
political environment. Formally: X = (2 (Z). Speciﬁc Z factors are introduced and
explained in Table 3.3, and inference is made about favored arrangements in the presence
of each factor. The framework is then used in the next section to show the degree to
which these factors affect the choice of organizational form in the cotton and tobacco

sectors in the Zambezi Valley of Mozambique.

 

'3 As show in Figure 3.5, spot marketing and full vertical integration are two extreme forms of
organization; contract farming can be seen as a continuum, tending to either of the extremes depending on
the terms agreed and the overall environment where it operates.

26

Figure 3.5

Organization of Production and Trade
A Conceptual Framework

Vertical Integration

 

Contract Farming
(Interlinked Transactions)

 

Spot Marketing

 

 

L
r

X

Y — Alternative Institutional Arrangements/Forms of Organization
X — Transaction characteristics: asset speciﬁcity, uncertainty, complexity
and frequency.

Table 3.3 summarizes how each factor affects transaction costs and its
implications for the type of institutional arrangement likely to result. In the next section
we apply the same framework to the cotton and tobacco sub-sectors in the Zambezi

Valley of Mozambique.

27

Table 3.3. Transaction Cost Factors and Institutional Arrangements

 

Type of Organizational Forms
Effect on Transactions Costs Most Favored
Factor Spot Contract Vertical
Markets Farming Integration
Production chagcteristics

High labor intensity Increases supervision costs and requires X X
capital saving/labor using technologies.
Economies of scale Requires high initial investment and high X

cash flow to be sustainable; generally not
feasible for smallholders.

 

High returns to inputs, complex Requires effective research and extension, as X X
management well as timelyavailability of inputs.

Marketing/processing characteristics
High economies of scale in Leads to the need for scale complementarity X X
processing that creates strong incentives for stable

supply of raw materials through more
coordinated arrangements.

High quality standards Increases returns to close vertical X X
coordination.
High perishability Increases the costs of not having a stable X X

market. Increases returns to close vertical
coordination.

High value to weight/volume Increases risk of large loss in farm to market X X
transaction.

Low value to weight/volume Increases unit transport costs. X

Principal market is export Tends to reduce number of buyers and risk of X X

default in CF; quality standards usually
higher; greater economies of scale.

 

 

 

 

 

Many potential buyers Increases cost and risk of default in CF. X X
Requires processing before ﬁnal sale Tends to reduce number of buyers and risk of X X
default in CF.
Exogenous economic & political factors
Land scarcity/high population Increases land cost, political difﬁculties X X
density obtaining large tracts.
Agriculture has a large share in the Increases land cost. political difﬁculties X X
labor force obtaining large tracts.
Endogenous economic ﬂolitical factors
Poorly integrated output markets Increases procurement costs and marketing X X
costs in general. Increases returns to
coordination.
Missing input/factor markets Non availability of necessary production X X

inputs limits reliance on spot markets and
increases the returns to vertical coordination.

Poor communications Raises cost of active vertical coordination, X X
especially contract negotiation and
enforcement.
Low literacy/educational levels Raises cost of ensuring adoption of new X X
among farmers production technologies/ management
practices; raises cost of collective action.
Weak property rights enforcement Increases uncertainty with regard to reliance X

in contracts and the use of collateral.
Increases the risk of default in CF.

Weak local government May make coordination more difficult; may X X
be easier to accumulate large tracts of land.

Source: Author's Conceptualization and Benﬁca, Tschirley, and Sambo (2002).

28

3.3.2. Contract Faming: Nature and Determining Factors

This section uses the analytical framework developed in the previous section to
look at the major factors determining the current form of organization of production and
trade in cotton and tobacco sectors in the Zambezi Valley of Mozambique. Note that
contract farming is the sole form of organization observed for both sub-sectors in the
region. Therefore, we will start by deﬁning that particular form of organization and

characterize it for the case of the cotton and tobacco sub-Sectors in the study region.

3.3.2.1. The Nature of Contract Farming Operations in Mozambique

In contract farming, farmers agree with processorS/traderS/exporters, through
formal or informal contracts, to limit their production and marketing behavior in return
for some level of service provision and purchase guarantee. These arrangements are best
viewed in the Zambezi Valley of Mozambique, and in many parts of the developing
world, as a response to missing institutions and widespread failure of input and credit
markets and to poor or absent service provision. In addition to processors’ need to ensure
sufﬁcient volume of purchases to reduce unit marketing and processing costs, concerns

about product quality often Signiﬁcantly affect the structure of these relationships.

Both cotton and tobacco schemes take the form of forward resource management
contracts. These contracts differ from the Simple sale/purchase contracts because they
include stipulations regarding the transfer and use of Speciﬁc resources and/or managerial
functions (Jaffee and Morton, 1995). Forward resource management contracts partially
internalize product and factor transactions, and are sometimes referred to as interlinked

contracts or interlinked markets (Minot, 1986; Glover and Kusterer, 1990; Dorward et a1,

29

1998). Given the current stage of development of rural agricultural input and credit
markets in the country, farmers have little access to those resources. The contracts,
designed to ﬁll that gap, consist in the ﬁrms supplying seeds and chemical inputs on
credit, along with technical assistance on Speciﬁc areas of land. Farmers agree to utilize
the inputs as instructed, and to sell all their production to the ﬁrms at harvest at pre-
determined prices. Input costs are deducted at the time of the harvest/marketing. Given
the lack of alternative contract enforcement mechanisms, the Mozambican Government
has granted the agro-industrial ﬁrms legal monopsony power over Speciﬁc geographic
areas, referred to as concessions. Farmers in these areas are not permitted to sell to any
but the concession holder. In the Zambezi Valley, production in these schemes takes

place entirely on land “owned” by the individual farmers. '4

3.3.2.2. Factors Leading to Interlinked Transactions

Many of the factors identiﬁed in Table 3.3 operate exactly the same way for both
sub-sectors. Indeed, they are both dependent on quality raw materials for processing.
High quality requires the use of on-farm chemical inputs and speciﬁc production
techniques. In a country that has a high degree of failure in output, input and credit
markets, and a poorly educated populace, reliance on spot markets for such crops is not
feasible, and some degree of non-market vertical coordination is called upon to support
and sustain these value chains. We now turn to an evaluation factor by factor. See Table
3.4 for the incidence of individual factors in each sector.

Production characteristics. Current crop production technology in these sectors is

 

'4 In some other parts of the country, however, ﬁrms can also use designated areas or blocks within their
own land concessions for that purpose (Strasberg, 1997). Note that in Mozambique land is ofﬁcially state
owned, therefore non-tradable. .

30

characterized by high labor intensity, no economies of scale and potentially high returns
to inputs. In principle, particularly in a plantation setting, high labor intensity leads to
high supervision costs in a principal agent setting. In addition, under a plantation
arrangement, ﬁrms would have to pay the legal minimum wage for agricultural workers,
generally set at a level higher than the informal wages paid to hired labor by smallholder
growers and, for some cases, even above the implicit wage earned by cash crop growers.
With contract farming labor, supervision is transferred to the household. Given the
relatively simple production technologies, the relatively high level of use of family labor
and low level of hiring in cotton, household level supervision costs are relatively low in
contract farming for that crop. Tobacco uses more complex production techniques and
more wage labor, but that wage labor tends to be relatively well trained, which reduces
supervision costs.‘5 Economies of scale, to be achieved and sustained, normally require
high investment and cash ﬂow which favors vertical integration.16 High returns to inputs
demand a great deal of detail in input use that requires some degree of coordination. In
the current stage of development of the Zambezi Valley of Mozambique, capital
constraints associated with a poorly developed marketing system for inputs, outputs and
credit, makes contract farming the most feasible alternative to deal with the factors
associated with crop production in these sub-sectors and the reduction in the resulting
levels of uncertainty. The realization of this potential depends on the technological
knowledge of farmers and on the level of coordination and organization of the out-grower

ﬁrm extension systems.

 

'5 Costs associated with labor supervision are more accentuated in a plantation setting relative to household
level management, as principal agent (manager-worker) problems, especially derived from adverse
selection and moral hazard are far more intense when all workers have to be hired and work for a wage.

‘6 Overtime, in the accumulation process, if returns are promising, contract farmers may have the incentive
to invest in technologies that allow for the achievement of those economies of scale.

31

Table 3.4. Incidence of Selected Factors in Cotton and Tobacco Sectors

 

Is the Factor Present?

 

 

 

 

 

Factor @gree of incidence: +++, ++, +, -)
Cotton Sector Tobacco Sector
Production c1ga_cteristics
High labor intensity + + + +
Economies of scale + +
High returns to inputs, complex + + + + +
management
Marketing/processing chgracteristics
High economies of scale in processing + + + +
High quality standards + + + + +
High perishability - -
High value to weight/volume + + + + +
Principal market is export + + + + + +
Many potential buyers - -
Requires processing before ﬁnal sale + + + + +

 

Exogenous economic & political factors
Land scarcity/high population density - -

 

 

Agriculture has a large share in the labor + + + + + +
force

Endogenous economic & political factors
Poorly integrated output markets + + + ' + + +
Missing input/factor markets + + + + + +
Poor communications + + + +
Low literacy/educational levels among + + + +
farmers
Weak property rights enforcement + + + + + +
Weak local government + + + +

 

Legend: + + + Strong presence; + + Moderate presence; + Weak presence; - Absent.
Source: Author conceptualization.

Processing and marketing characteristics. There are a number of factors related
to processing and marketing/export characteristics that strongly favor the inter-linkage of
transactions in both sub-sectors. First, the two cotton ﬁrms in the region have a
considerable amount of unused processing capacity and therefore need more raw product
to exploit economies of scale in processing. In the tobacco sector all production is
currently exported raw, but getting volume is as well important to achieve economies of
scale in export. In both cases there is no competitive small scale processing option, so a
system that can ensure volume is needed. Second, in both sectors, more so in tobacco at

this point, the marketing system stresses quality standards and pays a premium for it. In

32

principle, returns to ﬁrms and farmers can be increased with further quality
differentiation. Since quality is strongly affected by how well and consistently production
is managed, coordination mechanisms through contract farming are necessary. Third, the
high value to weight/volume of these crops makes transport costs relatively cheap,
especially for tobacco. If the products were perishable, this would increase the risk of
high loss in farrn-to-market transactions under an independent production system and
favor a more coordinated approach. But both products are not perishable, so the high
value to weight ratio, although important, has little inﬂuence over the organizational form
governing transactions. Fourth, both crops are exported, which implies high returns to
product quality that means potentially high returns to more effective coordination. Fifth,
the presence of relatively few potential buyers and the need to process before ﬁnal sale
reduces the risk of default in contract farming. Without effective contract enforcement
mechanisms in place, however, contract farming can be jeopardized. In theory, these two
factors favor some form of non-market vertical coordination.

The bottom line is that dependence on quality output for processing that is highly
dependent on proper use of chemical inputs in an environment where input markets are
missing and human capital is relatively weak, makes reliance on spot markets infeasible.
Full vertical integration (plantation arrangements) could be considered but the labor
intensive nature of the production process makes labor supervision costs high. The
statutory agricultural minimum wage that ﬁrms would have to pay to workers can also be
a serious burden. This combination of factors renders some form of contract coordination
the most feasible alternative in both sectors.

While contract farming allows these systems to function, asymmetric information

33

and unbalanced bargaining power over issues such as prices and grading have emerged as
barriers tb its development. Furthermore, especially in cotton, low world prices, low
ginning outtums, and low productivity at the farm level have constituted major
constraints. These are, in part, consequences of a wide range of market and coordination
failures and weaknesses in the concession system, many of which are beyond the scope

of this study.

3.4. Farmer Selection/Performance and Effects of Participation

The previous section identiﬁed the factors leading to the dominance of contract
farming in the cotton and tobacco sectors in the Zambezi Valley. Contract farming can be
seen as a principal-agent game where a ﬁrm (the principal) works with a grower (the
agent) to produce a crop. In this process, the ﬁrm chooses the farmers with whom it
would like to contract and sets the contract terms. The ﬁrrn’s objective is to produce a
given quantity of output while minimizing direct and indirect (transaction) costs.
Farmers, in turn, will choose whether to participate. The combination of these choices
describes the selection process for the contract farming scheme (Warning and H00,
2000). The beneﬁts participants accrue will depend on the terms of the contract and their
own characteristics and endowments.

In this section, we assess the determinants of three related processes for each crop
study area: farmer participation in the production of the cash crop, participants'
performance with the crop, and whether farmer participation, independently or associated
to levels of land ownership and education attainment, has a signiﬁcant impact on crop
and total household income. We ﬁrst present the study sampling coverage and descriptive

statistics on scheme participants and non-participants in each concession area. Second,

34

we formally present the rationale for sample selection bias and the Heckman two-stage
sample selection bias correction model. Finally, we present the Farmer Scheme
Performance Sample Selection Model, and the Treatment Effects Model speciﬁcation,

estimation and results.

3.4.1. Study Area Sample and Comparison of Means

3.4.1.1. Survey Sample Coverage

The survey covered concession areas for four ﬁrms operating contract farming
schemes in the Zambezi Valley of Mozambique: both tobacco ﬁrms operating in Tete
Province - Mozambique Leaf Tobacco (MLT), and DIMON-Mozambique - and two
cotton companies, one operating in Tete Province (DUNAVANT-Mozambique) and the
other operating in Northern Sofala Province (C.N.A.). The survey targeted a total of 300
smallholder farmers, 180 in tobacco growing areas and 120 in cotton growing areas. Due
to sample attrition, the ﬁnal sample size for analysis was reduced to 276 observations;
159 smallholders for tobacco concession areas and 117 smallholders for cotton
concession areas. In both areas the sample comprised both grower and non-grower

smallholder households. More details on survey sampling are presented in Chapter 2.

3.4.1.2. Comparison of Means

We present two sets of descriptive statistics. A comparison of means for three
types of household level variables: demographic characteristics; farm assets and use of
hired labor; and levels of crop and household income (total and per capita) along with
prevalence and levels of selected income components. We also analyze how the outcome

variables - cash crop proﬁts, crop and total household incomes - vary across types of

35

growers, by land holdings and education attainment levels.

Results for the tobacco sector in Table 3.5 indicate that participants and non-
participants are not statistically different in terms of demographic characteristics such as
household size, labor endowments, education, and age of the head. Household headship is
almost statistically signiﬁcant with the likelihood for female headship relatively higher
among non-growers. Also, differences are not statistically signiﬁcant for the use of
animal traction, and the rate of diversiﬁcation into livestock and self-employment non-
farm activities. Statistically signiﬁcant differences exist for a number of variables. First,
total area owned, and total and per capita crop incomes, with growers having areas and
agricultural incomes signiﬁcantly larger, especially due to the cash crop. Note that net
proﬁts from tobacco average $730.74, i.e., 46.5% of total net agricultural income for that
group. About 30% of tobacco farmers lost money during the survey year.l7 Second, as
expected, tobacco growers own greater values of agricultural and marketing equipment
(hand tools and other equipment, including bicycles). Third, tobacco growers are twice as
likely to hire permanent labor.

Finally, non-growers have wage labor and non-farm self-employment incomes
that are much higher in magnitude than those for tobacco growers, but those differences
are not statistically signiﬁcant. These differences partially compensate for the large
difference in crop incomes and make the differences in total household income less
accentuated, but still signiﬁcantly different in a statistical sense. It is consistently
observed in Table 3.7 that tobacco proﬁts among participating farmers increase with land

holdings. The same pattern is observed for the relationship between net total crop and

 

'7 Results by ﬁrm indicate that the proportion was higher in MLT areas, about 36%, against 23% in
DIMON-Mozambique areas.

36

household incomes across land holding quartiles for growers and non-growers in tobacco
areas. Results with respect to education attainment are not so robust (Table 3.8). While it
appears that cash crop proﬁts are positively related to education, those differences vanish
as one considers total crop and household income in tobacco areas.

Table 3.5. Comparison of Mean Values for Selected Variables: Tobacco Growers

and Non-Growers

 

 

 

Type of Farmers Statistical Signiﬁcance of the
(mean values) Difference
Selected Variables Tobacco Non- t-Stat P > I t | LS of the
Contract Tobacco Difference 1/
Growers Growers
Demographic Characteristics
Female Headed Households (%) 5.13 11.90 - 1.49 0.14
Education of the HH Head (years) 3.22 2.76 1.00 0.32
Age of the Household Head (years) 38.50 40.52 - 0.95 0.34
Labor Adult Equivalents 3.45 3.68 - 0.88 0.38
Farm Assets
Total Area — hectares 6.94 4.36 2.84 0.01 **
Value of Manual Tools ($US) 28.63 15.59 2.16 0.03 "‘
Value of Equipment ($US) 66.60 36.63 2.58 0.01 **
Use of Animal Traction (%) 7.69 4.76 0.64 0.52
Use of Hired Labor
Permanent Labor (% using) 71.79 30.95 4.98 0.00 **
Income Diversification (%) .
Livestock 93.98 96.15 - 0.44 0.66
Self-employment , 60.15 53.85 0.56 0.55
Wage Labor Employment 24.81 53.84 - 3.03 0.00 **
Household Income (3 US)
Net Household Income 1,815.28 1,022.48 2.35 0.02 *
Net Household Income per capita 318.06 174.70 2.36 0.02 "
Net Agricultural Income 2’ 1,572.70 595.47 3.1 1 0.00 **
Net Agricultural Income per capita 274.23 98.26 3.18 0.00 "
Wage Labor Income 80.76 122.35 - 0.92 0.36
Self-employment (non-agricultural) 90.24 185.90 - 1.14 0.26
Livestock Income 90.1 1 79.50 0.35 0.73
Number of observations 1 17 42

 

“ Levels of Signiﬁcance (LS): I at the 10-percent level, " at the 5-percent level, “ at the l-percent level.
2’ Net revenues from tobacco sales averages $730.74 among growers, i.e., 46.5% of net agricultural income.

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

37

The comparison of means for the cotton sector is presented in Table 3.6. Results
indicate that among demographic variables only that for educational attainment of the
household head is statistically different between the two groups. 18 Results suggest that
non-grower smallholder households have higher formal educational attainment.
Regarding farm asset variables, total area is the only variable that shows a statistical
difference between growers and non-growers. However, due to the lower return to cotton
and the fact that non-growers plant more maize and other crops, the difference does not
translate into statistically signiﬁcant differences in total and per capita crop incomes
between the two groups. Net proﬁts from cotton average $93.60, i.e., only 18% of net
agricultural income for that group. These average proﬁts represent only 13% of those
obtained by tobacco growers. However, contrasted to tobacco areas, where close to one
third of the growers lost money, only about 20% of cotton farmers lost money during the
survey year.19 Growers have on average more physical and livestock assets than non-

growers, but the differences are not statistically Signiﬁcant.

The use of permanent labor is generally limited in cotton growing areas, but, as
expected, grower households are more likely to use that type of workers. However,
differences between those two groups are not statistically signiﬁcant. Estimated total and
per capita income is higher for growers, but again, the difference is not statistically
signiﬁcant at 10% or lower levels. Likewise, non-growers appear to have off-farm

incomes (both wage labor and self-employment) higher than growers.

 

'8 The household age variable is close to signiﬁcant, indicating that grower household heads tend to be
older than those in non-growing households.

'9 Results by ﬁrm indicate that the proportions were 19% among C.N.A. farmers, and about 21% among
Dunavant Mozambique farmers.

38

Table 3.6. Comparison of Mean Values for Selected Variables: Cotton Growers and
Non-Growers

 

 

Type of Farmers Statistical Signiﬁcance of the
(mean values) Difference
Selected Variables Cotton Non- t-Stat P > It | LS of the
Contract Cotton Difference "

Growers Growers

 

Demographic Characteristics

Female Headed Households (%) 5.74 6.67 - 0.18 0.86
Education of the HH Head (years) 2.60 3.40 - 1.88 0.06 +
Age of the Household Head (years) 44.33 40.40 1.46 0.15

Labor Adult Equivalents 3 .51 3.25 0.86 0.39

Farm Assets

Total Area (hectares) 3.97 2.81 2.58 0.01 **
Value of Manual Tools ($US) 12.94 1 1.81 0.61 0.54

Value of Equipment ($US) 46.75 33.44 0.97 0.33

Use of Animal Traction (%) 5.75 6.67 - 0.18 0.86

Use of Hired Labor

Permanent Labor (% using) 9.20 3.33 1.03 0.30
Income Diversification - Percent

Livestock 90.1 1 88.46 0.24 0.81
Self-employment ' 62.64 84.62 - 2.13 0.04 *
Wage Labor Employment 34.07 38.46 - 0.41 0.68
Household Income - $ US

Net Household Income 732.40 574.64 1.21 0.22

Net Household Income per capita 124.89 108.57 0.76 0.45

Net Agricultural Income 2’ 518.24 364.50 1.50 0.14

Net Agricultural Income per capita 86.69 65.85 1.13 0.26

Wage Labor Income 42.18 80.57 - 1.24 0.22
Self-employment (non-agricultural) 32.15 , 56.02 - 0.60 0.55
Livestock Income 85.37 72.29 0.61 0.54

Number of observations 87 30

 

" Levels of Signiﬁcance (LS): I at the lO-percent level, "‘ at the 5-percent level, " at the l-percent level. 2’
Net revenues from cotton sales averages $93.60 among growers, i.e., 18.1% of net agricultural income.

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

These higher off-farm incomes for non-cotton growers and the resulting lack of
signiﬁcant difference in total incomes are consistent with previous studies in
Mozambique (Tschirley and Weber, 1994; Tschirley and Benﬁca, 2001; and Walker et

al. , 2004). In addition, mean values for wage labor and self employment income in Table

39

3.6 indicate that non—growers tend to have higher incomes from these sources but the

difference is not statistically signiﬁcant. One may argue that these results suggest a

relative degree of stagnation in these economics; the cotton incomes are not yet capable

of pulling the rest of the economy into a dynamic mode.

Results in Table 3.7 indicate a positive but weak association between land

holdings and proﬁts from cotton sales. The relationship is more accentuated when it

comes to overall crop and household incomes, especially among non-cash crop growers.

On average, returns to education (Table 3.8) appear much less important for cotton

proﬁts than for tobacco. The econometric analysis will shed some more light on the

signiﬁcance of these indicative relationships.

Table 3.7. Mean Proﬁts, Net Crop andTotal Income by Land Area Quartiles

 

 

 

 

 

 

 

Mean . Cash Total Net Crop Income Total Net Household Income
Quartiles of Land Crop ($US) ($US)
Land Area Area Proﬁts Growers Non- All Growers Non- All
(ha) ($US) Growers Farmers Growers Partners
Tobacco Areas
Quartile 1 2.28 184.7 562.0 175.6 417.1 726.7 533.5 654.2
Quartile 2 3.92 41 1.8 749.7 507.5 689.2 947.7 649.1 873.0
Quartile 3 6.12 462.1 1,499.0 846.8 1,298.3 1,684.1 1,227.1 1,543.5
Quartile 4 12.71 1,601.4 3,056.9 990.8 2,798.6 3,437.7 1,246.3 3,163.8
Total 6.26 730.7 1,572.7 543.5 1,300.8 1,815.3 844.0 1,558.7
Cotton Areas
Quartile 1 1.72 76.6 302.5 231.3 271.7 458.4 381.8 425.2
Quartile 2 2.88 52.1 477.4 321.2 439.7 687.1 565.8 657.8
Quartile 3 3.70 81.4 486.3 418.1 474.5 747.0 720.8 742.5
Quartile 4 6.45 156.0 740.4 1,283.2 834.0 953.5 1,460.4 1,040.9
Total 3.67 93.6 518.2 458.7 503.0 732.4 661.0 714.1

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

40

Table 3.8. Mean Profits, Net Crop and Total Income by Education Attainment Level

 

 

 

 

 

 

Cash Total Net Crop Income Total Net Household Income
Years of Crop ($US) ($US)
Education of Proﬁts Growers Non- All Growers Non- All
HH Head ($US) Growers Farmers Growers Farmers
Tobacco Areas
No schooling 554.9 1,472.4 377.0 1,126.5 1,594.4 659.29 1,299.1
1 — 3 years 731.5 1,695.0 669.4 1,446.9 1,931.2 887.35 1,678.7
4 + years 833.9 1,501.3 550.7 1,259.6 1,821.9 948.50 1,599.9
Total 730.7 1,572.7 543.5 1,300.8 1,815.3 844.03 1,558.7
Cotton Areas
No schooling 65.9 652.5 21 1.0 564.2 1,005.6 483.3 901.1
1 - 3 years 90.4 426.6 342.5 408.8 519.3 508.0 516.7
4 + years 119.9 559.6 638.5 587.2 858.4 844.7 853.6
Total 93.6 518.2 458.7 503.0 732.4 661.0 714.1

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

In Figures 3.6 and 3.7, we explore how area planted with cash crops relates to
total area owned, by plotting both variables by total land area quartiles among cotton and

tobacco growers.

Figure 3.6

Cotton Area Cultivated and Total Area Owned
by Quartiles of Grower Total Area

1

 

I

 

 

 

I

 

 

7//////n

 

 

 

 

7/

I
4th Quartile

 

Area (hectares)
O N A O) on
_I_

 

N

1st Quartile 2nd Quartile 3rd Quartile
Quartiles of Grower Total Area

 

Area with Cotton Total Area

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

41

Figure 3.7

Tobacco Area Cultivated and Total Area Owned
by Quartiles of Grower Total Area

 

 

 

 

Area (hectares)
7/////,

 

 

 

 

 

/
0,g_mmrmm,m§ Q

‘7

1st Quartile 2nd Quartile 3rd Quartile 4th Quartile
Quartiles of Grower Total Area

 

Area with Tobacco Total Area

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

In both cases, it is clear that area planted with the cash crop increases with total
area owned. Correlation between the two variables is 0.76 among cotton growers and
0.92 among tobacco growers. It is worth noting that among cotton growers, the share of
area planted with cotton in total area (across quartiles) is relatively ﬂat; 38% (Quartile 1),
31% (Quartile 2), 37% (Quartile 3), and 39% (Quartile 4). Among tobacco growers those
shares increase signiﬁcantly between the lowest and the highest total area quartiles; 37%

(Quartile 1), 37% (Quartile 2), 42% (Quartile 3), and 54% (Quartile 4).

3.4.2. Conceptual Framework

In this section we wish to explain what determines the levels of cash crop income
(proﬁts or losses) of participants and whether participation in contract farming schemes
affects differences in total crop and household income between growers and non-growers
in tobacco and cotton growing areas of the Zambezi Valley, controlling for land

endowments and education attainment threshold effects, as well as demographics,

42

 

technology and location ﬁxed effects. Because cash crop income is only observed for a
sub-set of the population, we run into a sample selection problem usually referred to as
incidental truncation, i.e., the observation of cash crop income depends on another
variable, in this case, the participation in contract farming schemes.

To accurately estimate the determinants of cash crop income, and the effects of
participation on total agricultural and household income, we have to account for the fact
that there may be unobservable factors that affect both the likelihood of participation in
the schemes and the performance of participating farmers (Greene, 2003; Warning and
Key, 2002). To control and correct for this possible sample selection bias, we use a
standard sample selection model to access participation and within scheme performance,
and a selection adjusted treatment effects mOdel to assess if and how participation affects

total crop and household income levels.

3.4.2.1. Determinants of Cash Crop Income

(a) Sample Selection Bias and Correction
We now present a standard sample selection model that explains and addresses

the sample selection problem. Let the equation that determines sample selection be
Ci=}Zi+€i , E(e|z)=0 (3.1)
and the equation of primary interest be
y,- = ﬁx,- + u,- , E(u | x) = 0 (3.2)
where c,- is a dummy for participation, zi is a vector of variables thought to affect the
participation decision, yi indicates the level of outcome of participants, x,- is a vector of

variables assumed to affect the outcome, and e,- and u,- are disturbance terms. The model

43

assumes ui~N(0, o), e,- ~N(0, 1), and corr(u,- ,e,-)=p.

Several assumptions are made. First, we assume that the elements in x and z are
always observed. Second, we assume that (in addition to x) z is exogenous in (3.2), i.e.,
E (ulx,z)=0. Third, we require that x be a strict subset of z, with some elements of 2 not
included in x (exclusion restrictions). Fourth, Since the error term in the sample equation,

ei, is assumed to be independent of z, and x is a subset of 2, then e; is also independent

of x. Finally, we assume that e,- has a standard normal distribution (Wooldridge, 1999

and 2000).

Where does the bias come from? Correlation between the error terms “1' and e,-
causes the sample selection bias. To see how, let us assume that (u,- , e,- ) are independent
of 2. By taking the conditional expectation of (3.2) on 2 and e,- and considering that x is a
subset of 2, we have

£00 I 2139i): ﬂxr' + E(ui l ziaei) = ﬂxi + E(ui I er) (33)

Note that E(yi | zi,e,-) = E(u,- | e) because (”i , e,) is independent of 2. It follows
that if “i and e,- are jointly distributed with zero mean, then E( u,- |e,- )=p e,- for some
parameter p, and replacing this in (3.3) we have

EU; I 4991'): .Bxi + Per (34)

Although we do not observe e,- , we can use this to compute E ( y,- l z,~,c,-) , for
0,- =1, and get

E(yi l Ziaci)= ﬁx: +PE(er' | zisci) (35)

Taking into account the relation between c,- and e,- from Equation (3.1) and the

44

fact that ei has a standard normal distribution, it can be shown that E (el- | zi,c,-) is simply
the non-selection hazard, what Heckman (1979) referred to as the Inverse Mills' Ratio
(IMR), 202,), when e,- =1.

Thus:

E(y,- 121,6: =1)= ﬂxr + 91(7er (3.6)

Equation (3.6) indicates that the expected value of the outcome ( y,- ), given the set
of characteristics 2; , and the observance of y,- when c,- =l (i.e., the household engages in
contract farming) is equal to ,Bxi plus the IMR evaluated at m. The equation indicates

that we can estimate the parameters of interest, B's, using only the selected sample, and
that we should include Mm) as an additional regressor.

The parameter p deﬁnes the selection bias. If p=0, OLS of y on x using the
selected sample gives consistent estimates of [3. Otherwise, if p¢0, we have omitted a

variable that is correlated with xi. That is why Heckman (1979) points out that the

presence of the selection bias can be viewed as an omitted variable problem in the

selected sample. The parameter p will be equal to zero when u,- and e,- are uncorrelated.

(b) Specification with Sample Selection Correction and Threshold Effects

Since 7 is unknown, we cannot evaluate Mm) for each observation. The

following is a summary of the procedure used in this paper, also known as the Heckit

Method, named after the work of James Heckman”.

 

2° James J. Heckman received the Nobel Price in Economics in 2000 for his development of theory and
methods for analyzing selective samples. htmz/mobelprize.org/economics/lgureates/2000/indeX.htm1.

45

The ﬁrst step uses all the observations in the sample to estimate the Probit Model
of Ci on 2,- ,

Pr(c, :11 2,): CD(;z,-) (3.7)

Equation (3.7) returns the estimates of y, i.e., the determinants of participation in
contract farming. The inverse mills ratio, IMR (A), is obtained from these estimates for
each observation i, as 2.,- = ¢(7r:,-)/ 002,-), where ¢(}z,-) and <1>(}2,-) are the normal
density and distribution functions, respectively.

The second step consists in running an Ordinary Least Squares (OLS) Regression.
The Net Cash Crop Income Determinants Model uses the selected sample, i.e.,

observations for which c,- =1, to run

4 3
yr = 209/117+ Z5£Ekr+ﬂxr' +Pllﬂi)+ui (3-8)
j=2 k=2

where, yi is the net cash crop income, A ji are owned land area quartiles, Eki are
education attainment dummies, and x,- (other demographic, assets, technology and

locational factors) is a subset of z,- from the ﬁrst stage.21 Equation 3.8 returns estimates

of the determinants of net cash crop income, a? , 60 , and 13's, and the sample selection

bias coefﬁcient p.

 

21 . . .
Elements excluded from 2, are known as exclusron restrlctlons.

46

3.4.2.2. Effects of Contract Farming on Crop and Household Income

(a) The Treatment Effects Model
The treatment effects model is an extension of the selectivity model presented in
the previous section. It estimates the effect of an endogenous binary treatment on a

continuous fully observed variable, conditional on the independent variables. In our case,

it is the effects on total crop or household income ( yi ) of participation in contract
farming operations (c,- ). The primary regression of interest is
yr = ﬂxr' + (Pct + er (39)
where, ci is a binary decision variable, that stems from an unobservable latent
variable that is assumed to be a linear function of the exogenOus covariates and w,- and a

random component u,-. Speciﬁcally,

rlr

c,- = 7w,- +u,- (3.10)
The decision to obtain the treatment (participate in contract farming) is made

according to the rule
. It
Ci = I If Ci > 0
c,- = 0 , otherwise

where e,- and u,- are bivariate normal with mean zero and covariate matrix

Cov(e,-,u,-) = [2 p]

1

This model has many versions and has been applied in a variety of contexts

(Bamow et al., 1981; Maddala, 1983; Angrist, 2001; and Greene, 2003). The model is

47

estimated either by maximum likelihood (MLE) or through a two-step procedure. The
MLE estimation can be time consuming with large datasets and the two-step estimation
with consistent covariance estimates provides a good alternative (StataCorp, 2003).

In the ﬁst stage of the two-step option (Maddala, 1983), one obtains the probit
estimates of the treatment equation

Mo =1 I Wt): ¢(rwi) (3.11)

From these estimates, the hazard, h,- , for each observation i is computed as

hr- =¢(W)/¢(W) if ct =1 . and

11i = ¢(rWi)/ [1 - ¢(7Wi)l if Ci = 0
where h,- = ¢(7w,-) and <D(7w,-) are respectively the density and distribution functions of
the standard normal evaluated at w.

By taking the difference in the expected outcome between participants and non-
participants in this model,
Eb’i 1 Ci = laxisWiI— EIVi | Ci = OaxiiwiI = (0 + P[¢(7Wi)/¢(7Wixl - ¢(7Wi»], it becomes
clear that if the selectivity correction is omitted from the second step equation, the OLS
will overestimate the effect of the treatment (Greene, 2003).

(b) Speciﬁcation with Treatment and Threshold Effects

The Threshold Treatment Effects Model of Contract Farming on Crop and

Household Income uses the full sample to run

4 4 3 3
0 0
yr =ﬂxi+¢ci+ XZajAji+ ZzajciAji+kz5kEki+ ZagciEki+PhiIWi)+ui (3-12)
1: j: =2 k=2

where, yi is total crop or total household income, c,- is the participation dummy, A ji

48

refers to land holdings quartiles, E ki are education attainment dummies, and .h, is the

sample selection hazard variable. Both the land holdings and the schooling variables

(A ji and Eki) are interacted with the participation dummy (of) to assess the effects of

participation associated with land and education thresholds. The model generates OLS
estimates of the average and threshold treatment-effects coefﬁcients (p, 01’s and 5’s, the
B's (effects of other variables), and the sample selection bias coefﬁcient p.

From the results of regressions (3.8) and (3.12), we can test for sample selection

bias using the t-statistic on it and h;, respectively, as a test of Ho: p=0. Under the null

hypothesis there is no sample selection bias.

3.4.3. Model Estimation and Discussion of Results

3.4.3.1. Farmer Participation and Performance in Contract Farming

The Farmer Selection/Participation Equation (Equation 3.7) is estimated using
the entire sample. The dependent variable is a dummy equal to 1 if the farmer participates
in the scheme and 0 otherwise. It is assumed that farmer likelihood of participation is
affected by four sets of factors: demographics, asset and factor endowments and
technology, income diversiﬁcation, and location. The variables associated with each
factor are as follows.

I Demographic Characteristics. The demographic variables include gender
and age of the household head, number of labor adult equivalents in the
household, and dummy variables for the level of formal education attained
by the household head Education dummies include “no schooling”

(excluded dummy), “1-3 years of schooling”, and “more than 3 years of

49

schooling”.

Household production assets and technology. These include quartile
dummies of total area ownedn, a dummy for the use of animal traction,
the value of hand tools, and the value of other agricultural/marketing
equipment, including bicycles.

Income diversification variables. It is hypothesized that households that
have signiﬁcant involvement in non-cash cropping activities will, given
the constraints in labor and other endowments, be less likely to enter into
cash crop contracts. To account and test for that, we include dummy
variables for livestock, self-employment and wage labor activities.
Spatial/Location variables. These variables are district-Firm level ﬁxed
effects and are included to account for the differences across locations in
the level of development, including natural resource endowments, physical
and communications infra-structure development, and other factors. In
tobacco areas we dropped the district of Angénia (MLT Area) and in

cotton areas the district of Gorongosa (C .N.A. Area).

The Farmer Performance Equation (Equation 3.8) uses only those farmers that

participate in the contract farming schemes in each area. The dependent variable is the

net value of cash crop income, i.e., after deducting the value of inputs provided on credit

by the out-grower ﬁrm and wage labor costs.

The explanatory variables include all demographic variables, household

production assets and technology, all location variables as previously deﬁned, and

 

22 Average land area for each quartile — across all households - is presented in Table 3.7.

50

Lambda, 2.02,- ). If statistically signiﬁcant, Lambda indicates and corrects for the

presence of sample selection bias. All income diversiﬁcation dummies were treated as
exclusion restrictions, variables contained in the selection equation but assumed not to
affect scheme performance.23 Model results are presented in Tables 3.9 and 3.10 for
tobacco and cotton areas, respectively. We discuss the results for each area separately in
(a) and (b) below.

(a) Results for Tobacco Farmers

Probit results for tobacco contract farming areas in Table 3.9 indicate that
household participation in tobacco contract farming schemes is more associated with
endowments, technology, and income diversiﬁcation opportunities than with household
demographic characteristics. While point estimates indicate that female headed
households are less likely to engage in tobacco production, the statistical signiﬁcance of
that result is not strong. Unexpectedly, results indicate (although without statistical
signiﬁcance at any relevant level) that households with more adult equivalents are less
likely to engage in the contracts. A similar result was found by Warning and Key (2002)
in their assessment of the Arachide de Bouche confectionary peanut program in Senegal.

The use of animal traction and the value of manual toOls are positively associated
with the likelihood of farmer participation in tobacco contract farming schemes. The
value of other equipment, including bicycles, has a small, positive, but statistically
insigniﬁcant effect. Also surprising, in light of the means comparisons in Table 3.5, is
that households with more land are not more likely than others to grow tobacco.

The probit results suggest that households that have access to alternative sources

 

23 We tested the statistical signiﬁcance of these variables and none were statistically signiﬁcant in
explaining the levels of cash crop income, but some were signiﬁcant in explaining participation.

51

of income are less likely to directly participate in contract farming schemes in tobacco
areas; households drawing income from livestock sales and wage labor are less likely to
get engaged in tobacco production. This suggests that if these options are sufﬁciently
proﬁtable households will invest more family resources in those. These may help
explaining the result on the negative relationship between available adult equivalents and
the likelihood of participation. Effectively smaller households may rely on permanent
wage labor and get engaged in tobacco while larger households may draw resources into
these activities, including in selling labor to smaller households that choose to grow
tobacco under contract. Education of the household head does not statistically increase
the likelihood of participation. The Model Pseudo-R2 is 0.25.

The analysis of the determinants of tobacco net cash income in the second step
does not indicate the presence of sample selection bias, i.e., the coefﬁcient of lambda is
not statistically signiﬁcant at 10% or less.

Once households choose to engage in tobacco, some effects are worth noting.
First, female headed households have mean net tobacco proﬁts $400 lower than their
male counterparts. Second, regarding farm endowments and technology, land has no
effect on net tobacco income until the fourth land area quartile, when it has a large and
highly signiﬁcant effect; while at lower levels of land holdings (Area_Q2 and Area Q3)
the differences are not statistically Si gniﬁcant, average proﬁts of land rich households
(Area_Q4) are $780 higher than that of their land poor counterparts (Area_Ql). The
value of manual tools also has a positive partial effect on net tobacco income, although

the effect of the variable is relatively small and only signiﬁcant at the 10% level.

52

Table 3.9. Determinants of Proﬁts from Tobacco Production
Parameter Estimates

 

 

 

 

 

 

Explanatory Variables 1“ Stage: Participation V 2"6 Stage: Net Income/Tobacco
Coeff 2 P > I z | Ls2 Coeff t-stat P ﬁg Ls2
Demographics 3’
Female headed household - 0.375 0.84 0.40 - 405.56 1.95 0.05 *
Age of household head - 0.013 0.89 0.38 - 5.44 0.82 0.42
Labor adult equivalents - 0.154 1.29 0.20 106.51 1.26 0.21
Education: 1-3 years - 0.071 0.20 0.84 - 148.86 0.66 0.51
Education >3 years 0.024 0.06 0.95 17.55 0.07 0.94
Assets and Technology 4’
Area_Q2 0.333 0.92 0.36 247.07 1.36 0.18
Area_Q3 0.027 0.06 0.95 - 78.32 0.34 0.74
Area_Q4 . 0.500 0.96 0.34 780.34 2.30 0.02 *
Use of Animal traction 1.198 2.35 0.02 * 198.83 0.48 0.63
Value of manual tools 0.023 1.70 0.09 + 8.47 1.79 0.08 +
Value of other equipment 0.004 1.22 0.22 3.86 1.51 0.13
Diversification Activities
Has livestock income - 1.026 1.90 0.06 +
Has Self-employment inc 0.257 0.89 0.37
Has wage labor income - 0.879 2.88 0.00 *
A gro-Ecological Effects
Mid-A [tirade
Macanga/MLT - 0.831 2.15 0.03 "‘ 30.78 0.10 0.92
Mualadzi/DIMON 0.161 0.43 0.67 83.19 0.41 0.69
AngOnIa/MLT(dropped)
Lower Altitude
Maravia/MLT - 0.361 .85 0.40 - 600.79 2.68 0.01 **
Luia/DIMON - 0.543 1.17 0.24 - 787.16 3.72 0.00 **
Inverse Mills Ratio (1 ) 229-53 1-03 0-31
Constant 1.544 1.85 0.07 + - 170.74 0.41 0.68
Number of observations 159 1 17
Wald ch12 (18), 45.25
Prob > ch12 0.0004
Pseudo R2 0.25
Log pseudo-likelihood — 81.62
F(16, 100) 4.12
Prob > F 0.0000
R — Squared 0.46
Root MSE 913.62

 

" Probit equation for participation, 1 if participates, 0 otherwise. 2’ Level of signiﬁcance (LS): I 10%, “ 5%,
** 1%. 3’No schooling (Education=0) is excluded. " Quartile 1 (Area_Ql) is excluded. Proﬁts and value
of assets are expressed in $US.

Source: Zambezi Valley Tobacco Concession Areas Study, 2004.

Third, there are no threshold effects of education on tobacco proﬁts. This is a

somewhat surprising result in a crop that is relatively intensive in management and

53

production speciﬁcity. We investigated: (a) if there were any statistically signiﬁcant
relationship between land area and education, and (b) if education determines the levels
of self-employment and wage labor income. Results indicate that land and education of
the household head have a weak correlation coefﬁcient of 0.11, and that, when running
the proﬁts determinants regression without the land area variables, the education variable
remains statistically insigniﬁcant. These results lend credence to our original ﬁnding that
returns to education are low, even in this demanding crop. Furthermore, a two stage
regression analysis on the determinants of off-farm income shows that education is an
important determinant of both self-employment and wage labor income in tobacco
growing areas, but only wage labor earnings (in the second stage) are statistically
increased with increased educational attainment of the head. For detailed regression
results, see Appendix B.

Fourth, agro-ecology matters. Results suggest that farmers operating in north high
altitude areas in Macanga (MLT) and Mualadzi (Dimon) have proﬁts pretty much in line
with those in Angonia (MLT), the omitted dummy, while those in Luia (Dimon) and
Maravia (MLT) in the lower and drier south have proﬁts statistically lower. For a
comparison of yields and proﬁts across ﬁrms, see the cumulative distribution functions in
Figures 3.8 and 3.9.24

Overall, the OLS model has a good explanatory power, R2 =0.46, and the F-test

and the probability value are also highly Signiﬁcant.

 

2‘ Data on farmer experience (number of years farmers have grown tobacco) were collected for the Dimon
area. In ﬁrm speciﬁc sample regressions, both the linear and the quadratic speciﬁcations did not show any
signiﬁcant effects.

54

Cumulative fraction of farmers

Figure 3.8

Cumulative Distributions of Tobacco Farmer Yields, by Firm
Zambezi Valley - Mozambique, 2003/2004

 

 

 

1 I 1 I
O 1&1)

2000
Tobaoo Famar Welds (kgs/Ia)

 

— M_T ----- Drrm_I\/bzarrbique

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

55

Figure 3.9

Cumulative Distributions of Tobacco Farmer Proﬁts, by Firm
Zambezi Valley - Mozambique, 2003/2004

 

 

 

 

 

 

 

/,_

9 no u
m o
E
.59
1....
o co ._
c o
.9
*6
‘3
U—
(I) V .4
.2 '
E
:3
E
O N- ‘

o

I I I I
—2000 0 2000 4000
Tobaoo Farmer Proﬁts (USD)
NLT ----- Dirmn__Mozarrbique

 

 

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

(b) Results for Cotton Farmers

Results for the sample selection model for cotton areas are presented in Table
3.10. The probit model in the ﬁrst stage has a Pseudo — R2 of 0.23. Model results indicate
that choice of participation is inversely related to household head’s education. This result
is consistent with ﬁndings in other cotton growing areas of Mozambique where more

educated farmers tend to choose off-farm work over cotton. In sharp contrast with results

56

in tobacco areas, households with larger land areas than the base group (land poor) are
more likely to seek a contract in cotton; all the land holding threshold variables are
positive and statistically signiﬁcant.25.The difference between the two sectors with this
respect may be explained by the fact that, under current technological packages, in
addition to land, participation in tobacco also requires the (unobserved) ability to manage
production resources in a more complex set of ﬁeld activities than what is required in
cotton; in cotton land is the single most important resource.

Consistent with expectations, access to alternative sources of income reduces the
demand for cotton production contracts. Livestock income opportunities and self-
employment are negatively correlated With the likelihOod of participation in contract
farming; households do not appear to choose cotton if they have other good alternatives.26
Expected low yields and cash returns in cotton production at this point contribute to this
result. Note that even in C.N.A. areas where yields are considerably higher (see Table
3.2), cash returns are reduced by a more expensive input package and lower producer
prices; the latter persistently set at the ofﬁcial minimum in recent years.

The regression results for farmer performance in the schemes Show a p-value of
0.000 for the F-test of joint signiﬁcance and an Adjusted R2 of 0.64. The results indicate
that the coefﬁcient associated with the IMR is statistically signiﬁcant at 5%, which
indicates the presence of sample selection bias. Among the demographic variables, only

the labor adult equivalents variable is statistically signiﬁcant at 10%, indicating that once

in the schemes, additional adults generate positive returns to cotton proﬁtability; each

 

2’ The similarity in magnitude in the coefﬁcients of land thresholds Q2 to Q4 suggests that the threshold for
getting into cotton lies around the 25'h percentile, i.e., the greater difference is between the smallest 25%
and everyone else.

26 Unlike in tobacco areas where labor demand is more pressing and labor markets are much more active,
wage labor income does not compete with direct participation in contract farming in cotton areas.

57

additional adult adds on average $37 per year. Though not statistically signiﬁcant, the
coefﬁcients on education achievement are positive. That is a rough indication that,

although highly educated heads tend not to participate in cotton farming, the ones that do

may be more likely to perform better than the less educated ones.27

Like in tobacco areas, no Signiﬁcant positive effects of total area owned on proﬁts are
observed until the fourth quartile; land rich cotton smallholders have proﬁts that are
about $150 higher than those of land poor cotton growers. This suggests that in order to

beneﬁt from the crop, smallholders need to be relatively large.

Furthermore, the value of production and marketing equipment is positively
associated with returns to cotton growing. Finally, the analysis of district-ﬁrm ﬁxed
effects indicates that, controlling for other factors, average proﬁts in all Dunavant areas
are statistically lower than those in Gorongosa (C.N.A.). Only farmers in Maringue
(C.N.A.) achieve higher proﬁts than those in Gorongosa.28 The cumulative distributions
on Figure 3.10 indicate that yields for the C.N.A. farmers stochastically dominate those
for the Dunavant farmers.

 

27 We tested and found that education and land holdings are uncorrelated. Also, returns to education in
cotton growing areas are more sizable, though also not statistically signiﬁcant, in non-farm self-
employment activities. See Appendix B for detailed regression results.

23 Information on the number of years farmers have grown cotton (experience) was collected for the
Dunavant area. Both the linear and the quadratic speciﬁcations did not reveal any signiﬁcant effects. Note
that, since Dunavant operations started only a few years ago, not much variation is observed in the sample.

58

Table 3.10. Determinants of Proﬁts from Cotton Production

 

Explanatory Variables

Parameter Estimates

 

1St Stage: Participation V

2'“1 Stage: Net Income from Cotton

 

 

 

 

Coeff z P > I z 1 Ls2 Coeff t-Stat P > | t 1 LS2

Demographics 3'

Female headed household - 0.594 0.95 0.34 13.140 0.15 0.88

Age of household head 0.002 0.16 0.87 - 1.070 0.61 0.55

Labor adult equivalents - 0.193 1.55 0.12 37.033 1.75 0.09 +

Education: 1-3 years - 0.141 0.37 0.71 41.620 0.77 0.44

Education >3 years - 1.079 2.54 0.01 ** 85.253 1.36 0.18
Assets and Technology 4’

Area_QZ 1.137 2.82 0.01 ** 8.700 0.16 0.87

Area_Q3 1.400 3.12 0.00 ** - 3.310 0.05 0.96

Area_Q4 1.212 2.36 0.02 "' 148.887 2.00 0.05 *

Use of Animal traction
Value of manual tools
Value of other equipment

Diversiﬁcation Activities
Has livestock income

Has Self-employment inc.

Has wage labor income

District F bred-Effects
Chit’lta/Dunavant
Chifunde/Dunavant
Moatize/Dunavant
Caia/C.N.A.
Maringue/C.N.A.

0.507 0.62 0.53

0.020 0.94 0.35
0.002 0.52 0.61
- 0.887 1.75 0.08 +
_- 1.104 3.11 0.00 "
0.045 0.14 0.88
- 0.111 0.20 0.84
- 1.085 1.98 0.05 "‘
0.101 0.22 0.83

0.229 0.48 0.63
-0.104 0.20 0.84

Gorongosa/C.N.A. (dropped)

Inverse Mills Ratio (2. )-

97.614 0.65 0.52
- 3.861 0.80 0.43
1.279 3.59 0.00 "'

-566.612 7.22 0.00 “
-385.921 3.95 0.00 **
-142.559 2.08 0.04 "'
-l44.355 2.49 0.02 "
18.698 0.26 0.79

-1 54.986 2.09 0.04 *

 

Constant 205.148 1.39 0.17
Number of observations 1 17 87

Wald chi2 (19) 33.16

Prob > chi2 0.02

Pseudo R2 0.23

Log pseudo-likelihood - 62.57

F(17,69) 11.14

Prob > F 0.00

R — Squared 0.64

Root MSE 193.90

 

" Probit equation for participation, 1 if participates, 0 otherwise. 2’ Level of signiﬁcance (LS): I 10%, "' 5%,
" 1%. 3’No schooling (Education=0) is excluded. 4’Quartile 1 (Area_Ql) is excluded. Proﬁts and the
value of assets are expressed in $US.

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

59

Figure 3.10

Cumulative Distributions of Cotton Farmer Yields, by Firm
Zambezi Valley — Mozambique, 2003/2004

.4 .6 .8
1 4 1

Cumulative fraction of farmers

.2
l

 

O
J—

 

I I I I I

 

O 500 1000 1500 2000
Cotton Famrer Weldsmgs/ha)
CNA ----- Dunavant_Mozambique

 

 

 

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

Yet Figure 3.11 shows that, up to the 20th percentile of proﬁts, C.N.A. farmers
lose more money than Dunavant farmers; after the 20th percentile C.N.A proﬁts are
higher, but not by nearly as much as yields. This pattern is particularly due to the higher

cost of the input package and the lower prices in the C.N.A. areas.

60

Figure 3.11

Cumulative Distributions of Cotton Farmer Proﬁts, by Firm
Zambezi valley — Mozambique, 2003/2004

.4 .6 .8
l l J

Cumulative fraction of farmers

.2
l

 

 

 

 

 

o .
I I I I
-500 0 500 1000
Cotton Farmer Proﬁts (USD)
CNA ----- Dunavant_Mozambique

 

 

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

3.4.3.2. The Effects of Contract Farming on Crop and Total Income

The objective of this model is to assess whether farmer participation in the
contract farming schemes in cotton and tobacco concession areas of the Zambezi Valley
of Mozambique signiﬁcantly explains differences in the level of crop income and total
income of rural households. As previously, we explore threshold effects of land holdings

and education.

61

In spite of indications in section 3.4.1 that agricultural and household income of
participants, more signiﬁcantly in tobacco areas, are higher than those of non-
participants, we cannot yet attribute that difference to their participation in the schemes;
our analysis needs to take into account the possibility that the households that do
participate in the schemes could have obtained higher incomes even if they had not
chosen to participate, i.e., there may be factors that affect both their likelihood of
participating in the schemes and their crop and household income levels. That said it is
clear that assessing the impact of scheme participation by simply regressing crop income
on the participation dummy variables using OLS could bias the estimate of the impact of
participation. In this model we consider two different OLS regressions in the second
stage: one for net crop income determinants and another for net total household income
determinants. The explanatory variables are similar in both regressions and include:

- Treatment dummy for scheme participation that takes the value 1 if the

farmer participates in the contract farming scheme, and 0, otherwise;

' Threshold eﬂects interaction terms between land holding quartiles and

participation, and educational attainment classes and participation;

' Spatial/Location variables - district dummies relevant for each area, and

interactions between those and individual farmer participation status;

' Selection hazard variable: hi(7w,-), generated from the ﬁrst stage probit

estimation, separately for participants and non-participants.

In addition to these variables, we also include variables to account for
demographic factors and household production/marketing assets and technology, as

deﬁned in the proﬁt determinants model.

62

(a) Results for Tobacco Areas

The results of the treatment effects model are analyzed through the OLS output
related to the second stage of the procedure applied to both agricultural income and total
household income.29 In tobacco growing areas, average total crop income at the
household level is approximately $1,315 ($1,573 among growers and $596 among non-
growers), while total household incomes (crop plus income from farm and non-farm
activities off the household farm) is $1,606 ($1,815 among growers and $1,023 among
non-growers). The same set of independent variables, as previously described, are used in
both regressions. As shown in Table 3.11, the models for tobacco areas ﬁt reasonably
well, with an Adjusted R2 of 0.44 for the crop income regression and 0.43 for the total
household income regression. Both regressions exhibit highly signiﬁcant F-tests.

Several results stand out. First, the models ﬁnd no returns to education in crop
income, regardless of a'household’s participation status in the contract farming schemes.
Education beyond three years does signiﬁcantly increase total household income of non-
growers, reﬂecting higher off-farm earnings, particularly from wage labor, of more
educated non-grower households; participation in contract farming by such households
almost entirely offsets this advantage, though this effect is not signiﬁcant. These results
are consistent with Walker et al’s (2004) national analysis, and with Tschirley and
Benﬁca (2002).

Second, two results stand out related to land holdings. First, participation in
contract farming has no impact on crop and total household income until the fourth

quartile, when its effect is very large; interaction effects of participation and land holding

 

29 Since we used exactly the same selection equation applied in the previous section, we are not
emphasizing the probit results here.

63

dummies are only statistically signiﬁcant and sizable at the fourth quartile ($1,306 for
crop income and $1,576 for total household income). This result suggests the presence of
important returns to tobacco production (at least within the land area sizes seen in this
sample), perhaps through more efﬁcient use of hired labor. If true, the result suggests the
possibility of substantial growth in coming years in the number of “emergent” or
commercial smallholder households, driven by proﬁt opportunities in tobacco. This class
of farmers has been conspicuously lacking in Mozambique to date (Walker et al, 2004).

The ready availability of experienced labor in the area may be a key factor driving
this result. Second, the relatively greater magnitude and signiﬁcance of the coefﬁcient on
the fourth land quartile variable in the total income regression as compared to the crop
income regression, suggests that even larger farmers appear to not be giving up on
proﬁtable off-farm income generating opportunities.

Third, female headed households earn lower crop incomes than their male-headed
counterparts ($488 less), but differences in total household incomes are negligible in
magnitude and not statistically signiﬁcant; this suggests that diversiﬁcation into off-farm
activities by female headed households reduces gender differentiation in incomes in those
areas. Ownership of equipment beyond hand tools appears to increase agricultural
incomes: though the coefﬁcient is not quite signiﬁcant in the agricultural income

regression, it is signiﬁcant in the total income model and its magnitude is nearly identical.

64

Table 3.1]. Effects of Tobacco Contract Farming (CF) on Net Crop and Net Total
Household Income: Model with Land and Education Threshold Effects

 

OLS Parameter Estimates — Tobacco Areas

 

 

 

 

Explanatory Net Total Agricultural Income Net Total Household Income
Variables “ Coef. Robust p |Z|>z LsT Coef. Robust P |Z|>z 1.32
SE. SE.
Participates in CF 407.70 555.62 0.46 85.87 568.47 0.88
Demographics
Female head househ. - 488.01 239.68 0.04 * 0.66 282.52 0.99
Age ofhouseh. head 4.85 10.32 0.64 15.85 11.04 0.15
Labor Adult equival. 25.44 98.06 0.80 - 3.99 105.43 0.97
Education Threshold Effects 3
Education: 1-3 years 195.32 258.15 0.45 269.76 259.28 0.30
Education >3 years 361.14 312.48 0.25 718.92 320.28 0.03 *
[Education : 1-3]*CF - 482.02 572.20 0.40 - 452.16 581.29 0.44
[Education >3]*CF - 637.32 581.68 0.28 - 703.27 585.63 0.23
Land Threshold Effects 4
Area_Q2 527.93 222.43 0.02 * 401.17 257.28 0.12
Area_Q3 665.13 331.93 0.05 * 820.94 279.98 0.00 "
Area_Q4 723.32 396.06 0.07 + 691.65 359.09 0.06 +
Area_QZ*CF - 129.33 349.50 0.71 4.26 377.02 0.99
Area_Q3*CF 166.40 553.41 0.76 - 18.28 517.81 0.97
Area_Q4*CF 1305.86 631.67 0.04 * 1,575.96 652.95 0.02 *
Assets and Technology
Use Animal traction - 56.43 601.06 0.93 - 275.33 620.81 0.66
Value oftools 8.59 9.14 0.35 5.72 8.82 0.52
Value of equipment 4.31 2.81 0.13 4.38 2.39 0.07 +
Use fertilizer in maize 12.99 250.38 0.96 - 22.13 244.14 0.93
A gro-Ecological Eﬁ'ects
Mid-altitude
Macanga/MLT 165.83 371.25 0.66 - 159.92 345.50 0.64
Mualadzi/DIMON 774.05 459.01 0.09 + 423.32 419.30 0.32
Angonia/MLT 224.71 341.65 0.51 - 91.76 283.13 0.75
Macanga/MLT‘CF 662.23 722.84 0.36 942.34 722.45 . 0.19
Muala/DIMON*CF 182.69 602.86 0.76 357.91 586.89 0.54
Angonia/MLT‘CF 141.48 553.88 0.80 265.72 545.30 0.63
Lower altitude
Maravia/MLT - 12.51 410.23 0.98 - 244.43 382.95 0.52
Maravia/MLT*CF 90.38 772.74 0.91 36.57 760.10 0.96
Luia/DIMON (excluded)
Select. hazard ratio (h) 331.11 246.49 0.18 68.56 242.59 0.78
Constant - 1,101.09 793.64 0.17 - 679.39 773.48 0.38
N 159 159
F (27, 131) 4.11 4.92
Prob > F 0.0000 0.000
R — Squared 0.44 0.43
Root MSE 1,207.00 1,258.10

 

”OLS regressors. ” Level of signiﬁcance (LS): * 10%, * 5%, H 1%. 3’No schooling (Education=0) is
excluded. " Quartile 1 (Area_Ql) is excluded. Crop income, total HH income, and the value of assets, are
expressed in SUS. Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

65

Table 3.12. F-Tests of Joint Significance of CF and Education and Land
Thresholds: Tobacco Areas '

 

 

 

 

Net Agricultural Income Regression Net Total Household Income
Regression
Combined F (2,131) Prob>F L Combined F(2,l31) Prob>F L
Effect S Effect S
1/ 1/
CF -Education Threshold Effects
CF&[Education: l-3]"‘CF (74) 0.40 0.67 (366) 0.40 0.67
CF &[Education >3]"‘CF (229) 0.61 0.54 (617) 0.92 0.40
CF -Land Threshold Eﬂects
CF&Area_QZ"'CF 279 0.33 0.72 90 0.01 0.99
CF&Area_Q3*CF 574 0.28 0.76 68 0.01 0.99
CF&Area_Q4"‘CF 1714 2.26 0.10 + 1662 2.91 0.05 "‘

 

" Level of signiﬁcance (LS): + 10%, * 5%, '” 1%. The effects on net agricultural and net total household
income are expressed in SUS. Source: Zambezi Valley Cotton and Tobacco Concession Areas Study
Survey, 2004.

District level ﬁxed effects are relatively weak. Results indicate that in terms of
crop incomes among all households only in Mualadzi (Dimon) they are statistically (and
in magnitude) higher than those in Luia (Dimon). There are no participation-location
effects in crop incomes, which suggest relatively balanced outcomes across participants
in different locations. The coefﬁcient of the selection hazard ratio is not statistically
signiﬁcant in either regression, which indicates that correction for the selection bias in
not important in this model.30

(b) Results for Cotton Areas

Results of the second stage OLS regressions for crop and household income for
cotton areas are presented in Table 3.13. In cotton growing areas, average household total
crop income is $479 ($518 for growers and $365 for non-growers), while total household

incomes (crop plus income from farm and non-farm activities off the household farm) is

$692 ($732 for growers and $574 for non-growers). F-tests of joint signiﬁcance are

 

3° In the analysis of the ARB Program, Warning and Key (2002) found a similar result regarding sample
selection bias.

66

highly signiﬁcant, with 11%. of 0.60 and 0.48.

The results indicate that none of the demographic variables are statistically
signiﬁcant in either regression. The coefﬁcient on the participation dummy — which
reﬂects returns to cotton farmers in the lowest land and education classes — is positive in
both cases, but is not statistically signiﬁcant. Likewise, none of the participation-land
thresholds are statistically signiﬁcant. This is somewhat consistent with earlier results
indicating relatively low productivity levels in cotton and some kind of food ﬁrst strategy
being carried out by both cotton growers and non-grower, with maize production clearly
competing for household labor and land resources; the end result is the prevalence of a
situation where crop and total household incomes between cotton growers and non-
growers, after controlling for demographic, factor and asset/technology endowments, and
spatial factors, are not signiﬁcantly different.

We also ﬁnd that the value of farm and marketing equipment (excepting manual
production tools) is positively associated with higher crop and total household incomes,
but the magnitude of the effect is small. As expected, the value of manual tools is
positively associated with crop income.

We know that cotton yields. are substantially higher in C .N.A. areas than in
Dunavant areas (Figure 3.8); on the other hand, Dunavant has been paying better prices
than C.N.A. Therefore, there is still plenty of room for improvement in both productivity

and pricing. There is no evidence of participation-district ﬁxed effects.

67

Table 3.13. Effects of Cotton Contract Farming (CF) on Net Crop and Total
Household Income: Model with Land and Education Threshold Effects

 

OLS Parameter Estimates — Cotton Areas

 

Explanatory Variables U Net Total Agricultural Income Net Total Household Income

 

 

 

Coef. Robust P |Z|>z LS2 Coef. Robust P IZJ>z LS2
SE. SE.

Participates in CF 60.39 223.12 0.79 200.72 280.63 0.48

Demographics

Female head household - 61.67 110.41 0.58 - 55.45 169.74 0.75

Age ofhousehold head 1.37 3.18 0.67 2.60 4.14 0.53
Labor Adult equivalent 1 1.02 34.71 0.75 - 5.25 40.83 0.90

Education Threshold Eﬂects 3
Education: 1-3 years 99.30 157.58 0.53 107.91 187.02 0.57
Education >3 years 51.55 172.07 0.77 210.47 200.74 0.30
[Education : l-3]*CF - 303.33 226.18 0.18 - 252.81 280.34 0.06 +
[Education >3]*CF - 257.31 280.30 0.36 - 436.22 399.59 0.28

Land Threshold Effects 4
Area_QZ 1 16.97 137.16 0.40 296.49 174.84 0.09 +
Area_Q3 121.68 195.63 0.54 159.1 1 203.43 0.44
Area_Q4 761.19 283.64 0.01 " 718.56 317.42 0.03 *
Area_QZ‘C F 203.75 199.97 0.31 40.01 246.23 0.87
Area_Q3‘CF 152.56 257.67 0.56 322.72 271.95 0.24
Area_Q4‘CF - 299.96 346.81 0.39 - 251.27 421.62 0.55

Assets and Technology
Use Animal traction - 241.37 241.81 0.32 - 21.69 299.43 0.94
Value oftools 16.62 8.61 0.06 + 9.46 8.20 0.25
Value of equipment 2.05 0.621 0.00 '” 1.81 0.96 0.06 +

District F t'xed-Eﬂects
Chil’lta/Dunavant 18.15 238.88 0.94 81.18 250.80 0.75
Chihta/Dunavant’CF - 388.82 329.28 0.24 - 14.51 388.62 0.97
Chifunde/Dunavant - 467.27 250.65 0.07 + - 336.54 255.26 0.19
Chifunde/Dunava*CF 17.70 282.44 0.95 - 22.32 349.12 0.95
Moatize (Dunavant) - 319.40 153.28 0.04 "' - 386.35 190.54 0.05 *
Moatize/Dunava’CF 136.96 199.60 0.49 309.95 236.99 0.20
Caia /C.N.A. - 154.97 169.11 0.36 113.87 242.03 0.64
Caia/C.N.A.) *CF 139.27 229.78 0.55 - 86.24 313.12 0.78
Maringue/C.N.A. 52.32 157.32 0.74 140.00 212.27 0.51
Maringue/C.N.A.*CF 28.79 222.47 0.90 30.20 283.82 0.92
Gorongosa/ON.A.(exc1uded)

Select. hazard ratio (h) 106.59 150.70 0.48 160.53 202.70 0.43

Constant 16.921 249.33 0.95 66.02 277.75 0.81

N 1 l7 1 17

F (28, 88) 11.18 5.00

Prob > F 0.00 0.00

R - Squared 0.60 0.48

Root MSE 389.03 500.91

 

l[OLS regressors. 2’ Level of signiﬁcance (LS): i 10%, "' 5%, " 1%. 3[No schooling (Education=0) is
excluded. 4’ Quartile 1 (Area_Ql) is excluded. Crop income, total HH income, and the value of assets, are
expressed in SUS. Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

68

The coefﬁcient of the selection hazard variable is not statistically signiﬁcant
indicating the absence of sample selection bias in this model. Table 3.14 shows the F-
tests of joint signiﬁcance that assess the combined effects of participation per land and
education thresholds. In all cases, the effects are not statistically signiﬁcant.

Table 3.14. F-Tests of Joint Signiﬁcance of CF and Education and Land
Thresholds: Cotton Areas

 

 

 

 

Net Agricultural Income Regression Net Total Household Income
Regression
Combined F(2, 88) Prob>F L Combined F(2, 88) Prob>F L
Effect S Effect S
I/ 1/
CF -Educati0n Threshold Effects
CF&[Education: 1-3]*CF (242.94) 1.24 0.30 (52.09) 2.12 0.13
CF&[Education >3]*CF (196.91) 0.45 0.64 (235.50) 0.60 0.55
CF -Land Threshold
Eﬂects
CF and Area_Q2‘CF 264.15 0.54 0.58 240.74 0.28 0.76
CF and Area_Q3*CF 212.95 0.30 0.74 523.45 1.21 0.30
CF and Area_Q4*CF (239.56) 0.44 0.65 (50.54) 0.48 0.62

 

'/ Level of signiﬁcance (LS): + 10%, * 5%, ” 1%. The effects on net agricultural and net total household
income are expressed in SUS.

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

3.5. Summary of Policy Implications

Tobacco and cotton concessions in the Zambezi Valley of Mozambique have
provided a secure source of cash income to the rural population in areas where alternative
income generating activities are limited. In this section we look at some key issues in
each sub-sector as they relate to the results of the analysis presented in this essay and
elaborate on its policy implications.

Key results and implication from the econometric analysis in the tobacco sector
relate to the impacts of education, land holdings, access to wage labor, issues related to

labor migration, and the effects of environmental and technological spillovers. First, the

69

lack of returns to education in a crop as demanding as tobacco is surprising. Perhaps the
best interpretation is that great scope remains for improving ﬁeld practices, yields, and
proﬁtability; as companies strengthen their extension efforts and more farmers have more
time to learn proper techniques, we expect more educated farmers to begin earning higher
returns from tobacco.

Second, results on land holding size and access to wage labor may tell an
interesting story. Tschirley and Benﬁca (2001) showed that those with wage labor
income, especially those at the high end of this market, tend to maintain such income for
long periods of time. Boughton et al. (2005) showed that most income growth throughout
the country over the past six years has come from off-farm incomes, especially wage
labor. The research in this essay shows that households with such income are less likely
to grow tobacco; households without such income are the ones taking advantage of the
tobacco opportunity. As a result, tobacco cultivation may reduce income inequality.
However, many smaller farmers earn negative proﬁts from tobacco, while larger farmers
tend to earn large positive proﬁts. Over time, this pattern could drive substantial
expansion in the number of “emergent” smallholder farmers in the area. Those left
behind will be the smaller farmers who also have little access to wage labor
opportunities.

Third, recent expansion in tobacco production has been possible due in part to the
readily available labor knowledgeable of tobacco cropping, especially in border areas.

Survey data suggest that two-thirds of the 61% of farmers with permanent workers

70

employ at least one worker labelled as Malawian.3 ' Many of these are former tobacco
smallholders in Malawi that ﬁnd the wage labor opportunities in Mozambique more
proﬁtable. The likelihood of a farmer employing this type of labor increases with area,
and the proﬁts and household income of farmers that hire Malawian labor tend to be
above those that do not do such hiring. Approximately 25% of total payments to wage
laborers in the area go to this group. In terms of our model, these patterns raise concerns
about possible consumption leakages; yet, over 75% of the so called “Malawians” report
spending 9-12 months working in Mozambique, which suggest that a great deal of their
annual consumption takes place in Mozambique. In practice, then, our results suggest two
things. First, income leakage is not likely to be a major problem. Second, availability of
Malawian labor is important to the growth of the sector. One policy implication is that
efforts to ensure that Malawi migrants gain some kind of permanent residency that leads
them to spend more time and resources in Mozambican territory can be helpful both to
feed expansion of the sector and to spread beneﬁts in the local economy.

Finally, technological and environmental spillovers in tobacco growing areas need
to be more closely examined. On the positive side, growers and non-growers both are far
more likely to apply fertilizer on food crops than are farmers in other areas of the
country; it is likely that the provision of fertilizer for tobacco has contributed to this
pattern, through a combination of some diversion to food crops32 and greater familiarity
with the input leading to greater use. On the negative side, the rate of tree cutting by

tobacco growers far surpasses the rate of planting (Benﬁca, et al., 2005). Long-term

 

3 ' Evidence suggests that this labor force is a mixture of returned refugees (established in Malawi during
the Mozambique Civil War), family members of these returnees, and a genuinely new generation of
Malawian migrant laborers.

32 We do not know how common such diversion is.

71

consequences could be quite negative if these trends are not halted. Speciﬁc actions to
contain or reverse the situation are, therefore, required.

In the cotton sector, results from our analysis are in line with several other
studies in the cotton sector that have emphasized low prices and poor productivity
at the farm level as factors leading to the stagnation of cotton farmer incomes in
Mozambique (World Bank, 2005; Tschirley et al., 2006); Mozambique pays the
lowest prices in Africa, and farmer yields are also among the lowest. Our study
clearly documents the low proﬁtability of the crop relative to tobacco in the
Zambezi Valley region. The concession model as applied in Mozambique, which
precludes competition among companies and does not balance this with any
effective performance monitoring system, must be considered an important
contributor to the problem of low prices and also low productivity; resolving this
problem through more sophisticated management of the concession system has to
be a high priority for the government and other stakeholders over the next few
years.

In this study we ﬁnd, through our econometric analysis, that increased proﬁts in
cotton can be achieved with increased farm size and a higher level of production assets.
However, there are no landholding threshold effects on income from all crops, nor on
total household income. This seems to suggest some stagnation in these economies,
where cotton activity is not yet capable of pulling the rest of the economy into a dynamic
mode through strong economic linkages. Again, these results are quite different from
those found for tobacco areas.

Improving the contribution of cotton to smallholder livelihoods in Mozambique

72

requires a host of improvements in the quality of the seed stock, systems for treating seed
prior to distribution to farmers, improved input packages linked to effective extension
with farmers, and improved pricing. The entrance of new companies into the sector such
as Dunavant — with a good productivity and pricing record in Zambia - and CNA, with
an impressive productivity record in Mozambique, holds the promise of improved results
for smallholder farmers. These results, however, have so far not been delivered for most
farmers. Reforms to the concession model currently in place would seem to be a
requirement for signiﬁcant future progress. Because many alternative approaches to
reform are possible, informed research on reform paths, linked to some kind of
participatory stakeholder process, should receive high priority among both public and

private groups interested in the sector, and in smallholder welfare.

73

CHAPTER 4
DISSERTATION ESSAY TWO

Income Poverty Effects of Expansion and Policies in Cash Cropping Economies in
Rural Mozambique: An Economy-wide Approach

4.1. Introduction

Econometric analysis in chapter three indicates that growth in cash cropping,
especially of tobacco, can potentially affect income differentiation between growers and
non-growers. Tobacco might reduce differentiation by providing high returns to
households not engaged in remunerative off-farm activities, and by providing wage labor
opportunities to more households; at the same time, it could contribute to differentiation
by driving the growth of a group of relatively large, “emergent” tobacco farmers. The
relative magnitude and direction of each of these effects depends to a great extent on the
nature of second round effects that result from employment, production and consumption
linkages in those economies. This is an issue we did not deal with in Chapter Three; it
will be central to this chapter.

In this essay we use an economy-wide framework to analyze changes in the levels
of income poverty of different household groups following certain exogenous shocks to
the cotton and tobacco sectors; these shocks include expansion of contract farming
through capital injections, technology improvements, changes in world market prices,
and changes in Government trade policies. These issues are addressed using a regional
agricultural CGE model in the tradition of Lofgren and Robinson (1999) and Taylor et al.

(1999) which we developed and calibrated with a regional Social Accounting Matrix

74

(SAM). The model focuses on the Zambezi Valley area, which has received much of the
new investment in these sectors over the past decade.

Macro Economy-wide models, normally designed to capture the second and
higher feedback effects of investment and policy changes, have been used to analyze a
wide range of issues in Mozambique, such as aid dependence, marketing and agricultural
technology (Tarp et al. , 2002), the effects of HIV/AIDS, and the effects of multilateral
trade agreements (Amdt, 2003; 2005). However, macro level economy-wide models,
including national CGE models, typically abstract from local economies, and do not
provide the detail needed to uncover the full impact of policy changes on rural
economies; this is especially true when agricultural households are engaged in a portfolio
of farm and non-farm activities (Taylor et al., 1999), and when the analysis of issues
and/or the impact of Shocks that are speciﬁc to a relatively small region.

The essay is organized as follows. Section 4.2 explains the relevance of the study
for pro-poor growth policy strategies. Section 4.3 presents the data and analytical
methods that include the Regional SAM, the Regional CGE model, a discussion on
interlinked transactions, and a conceptual framework for the Representative Household
Approach to Poverty Analysis used in this study. Using SAM and household survey data,
representative characteristics of the regional economy are presented in Section 4.4.
Section 4.5 focuses on the deﬁnition and set up of the various policy and
technology/investment options that form the basis for the analysis, and the presentation

and discussion of results. Section 4.6 closes with conclusions and policy implications.

75

4.2. Relevance for Pro-Poor Growth Policy

Poverty reduction policy statements in Mozambique (GOM-PRSP 2001, 2005)
recognize that agro-industrial investments and export oriented agriculture can play a role
in reducing rural poverty. What these statements lack is speciﬁc knowledge that would
help to sketch strategies to enhance the impact of these investments on rural poverty in a
consistent and sustainable way. Current agro-industrial and agricultural export oriented
investments in the country demonstrate various degrees of connectedness with rural
households. The total impact of such investments or policies designed to promote them
depend on the Size of the direct and indirect/induced effects in a particular region (Hazell,
1984; Rogers, 1986; Haggblade et al. 1987; and Delgado et al. 1998). Alternative supply
chain institutional arrangements, ranging from the reliance on independent producers in
spot markets, to various forms of alliances and contracting, to full vertical integration,
will generate various degrees of initial employment and income effects, depending on the
crop and supply chain in question.

An important question for policy makers refers to the magnitude of the secondary
effects from investments, i.e., the Size of production, consumption, and employment
linkages in the local economy. If those effects are expected to be strong, pro-poor
economic growth strategies need not be directly or exclusively oriented to the poor.
Furthermore, in an environment characterized by serious resource constraints, and
bearing in mind that it is normally more expensive to reach the poor than the relatively
more endowed, a growth strategy that emphasizes second round effects in rural poverty
reduction may, if these effects are large, be preferred to one that focuses solely on the

maximization of direct effects to the poor (Tschirley, 2002; Benﬁca et al. , 2002; Carrilho

76

et al., 2002).33 The validity of this conclusion, that poverty reduction may be achieved at
a lower cost when emphasis is given to second round effects, depends to a great extent on
the structure of the economy where growth takes place that inﬂuences the relative costs
and beneﬁts associated with alternative investments and policy options. The preferred
approach becomes, therefore, an empirical question. In this study we look at that question
using data from the Zambezi Valley region of Mozambique where cash crops are grown
under contract, and baseline data allows for stratiﬁcation of farming households and the

subsequent analysis of relative impacts of policy changes on income poverty levels.

4.3. Data and Analytical Methods

This section presents the data and analytical methods used in this study, including
the Regional SAM for the Zambezi Valley (ZVR-SAM), the Regional CGE model (ZVR-
CGE), including its standard structure, and a discussion on interlinked transactions in
selected cash cropping activity sectors. Then, it presents a conceptual framework for the
Representative Household Approach to Poverty/Inequality Analysis. Finally, it discusses
issues related to the design and implementation of policy Simulations with the CGE

1

model.

 

33 This is due to the greater difﬁculty in reaching the poor with messages on innovative technologies, but
more importantly due to the lesser ability of the poorest households to make the adjustments and bear the
risks needed to adopt new technologies. Indeed, rational policy makers attempting to reduce poverty in the
face of a budget constraint will rely more on indirect effects than they would if costs to reach all groups
were equivalent (Tschirley, 2002).

77

4.3.1. The Regional Social Accounting Matrix for the Zambezi Valley

4.3.1.1. Introduction

A SAM is both a data system and a conceptual framework useful for policy
analysis (Thorbecke, 1998).34 SAMS have been used to model diverse economies and
institutional structures within various geographical scopes. Initial applications were
mostly modeling macro level issues in a national accounting context; more recently the
framework has been adapted to study micro (villages and towns) and sub-national (or
regional) economies. The strengths of the SAM framework are in its ﬂexibility and
adaptability to model a variety of economic structures and institutional setups. As a data
system, the SAM is comprehensive and disaggregated, as it includes transactions among
sectors, factors and institutions; consistent in the sense that for every income there has to
be an equivalent expenditure; and complete in that both the sender and the receiver of
every transaction need to be identiﬁed. For a given year, the SAM provides a snapshot of
the structure of the economy under investigation: the structure of production, inter-
sectoral linkages, distribution of factor value added among socio-economic groups, and
their expenditure patterns. The SAM framework consists of a square matrix of double-
account, in which rows represent receipts (revenues) and columns represent expenditures
(payments). As a comprehensive, consistent, and complete accounting method, it

requires balancing of revenues and expenditures in all accounts.35

 

3‘ The genesis of the SAM dates back to Sir Richard Stone’s pioneering work on social accounts (1978), for
which he received the Nobel Prize in Economics in 1984
( http://nobe1prize.org/economics/laureates/ 1 984/index.htm1). Subsequently, Pyatt and Round (1979)
further formalized the SAM and showed how it could be used as a conceptual framework for policy and
pslanning purposes.

Issues regarding the process of constructing the regional SAM for the Zambezi Valley and the balancing
procedure used to ensure the required accounting consistency are discussed Chapter Two.

78 t

4.3.1.2. The Structure of the Regional SAM

The Zambezi Valley Regional SAM (ZVR-SAM)‘keeps all the features of a
standard SAM. Some important ones are: the inclusion of non-marketed home consumed
commodities by farm households; the explicit treatment of marketing costs; and the
separation between production activities and commodities that allows any activity to
produce multiple commodities and any commodity to be produced by multiple activities.
In addition, the SAM has two other distinctive characteristics. First, to account for the
diversity of rural production activities, demand patterns, technologies, and market
structures, the SAM is highly disaggregated. In total it includes 83 accounts. Second,
agricultural activities (farm types) are mapped with household types. This allows for
better integration and subsequent modeling in a CGE framework of production and
consumption decisions in a non-separable fashion with the relevant differentiation across
farm-household types and activities (Lofgren and Robinson, 1999; and Taylor et al.,
2005)

The 2003-2004 ZVR-SAM includes six major types of accounts: (i) activities; (ii)
commodities; (iii) commodity marketing costs; (iv) factors of production: labor, land, and
capital; (v) institutions: households, government, and rest of the world; and (vi) savings
and investment. A schematic view of the SAM is presented in Table 4.1.

The Production Activities accounts describe the value of commodity inputs and
the payments to factors used in the production process (columns) and domestic market
sales and home consumption of goods and services (rows). The Zambezi Valley SAM
includes 18 such accounts. There are Six accounts for agricultural farms, four of which

represent cash cropping farms: two tobacco farm types and two cotton farm types. The

79

remaining two represent non-growers of cash crops in the respective areas. Other
activities include ﬁshing, livestock, food processing, beverage processing, other
processing or manufacturing activities, trading services, Government services, and other
services. Finally, we include four marketing and exporting activities representing each of
the four ﬁrms that operate contract farming schemes, supplying inputs on credit and
buying back the output that is subsequently exported, after adding some value.

The Commodities accounts record the value of total supply, including the value of
domestic production marketed and imports after taxes and marketing margins (columns)
and total demand, including demand for intermediary inputs by activities, private and
government consumption of goods and services, investment demand, and exports (rows).
We have 45 commodity accounts. The major cash crops are represented by eight
accounts, reﬂecting high level of disaggregation with ﬁrm-speciﬁc raw tobacco accounts,
raw cotton accounts, and another similar set of accounts for packed/graded tobacco and
cotton.36 Other accounts represent agricultural raw and semi-processed food commodities
ranging from maize grain and meal to fruits and vegetables (eight), animals and animal
products (six), processed foods (Six), agricultural inputs (four), non-food items (nine),
and services, ranging from health and education to maintenance (four).

Marketing Costs (MO/Marketing margin accounts quantify the wedges between
production and market prices for domestically produced goods sold domestically,
between import border prices and domestic market prices for imported commodities, and
between domestic and export border prices for exported goods. In principle, they account

for both transportation and the intangible costs of doing business, including procurement,

 

36 Note that packed tobacco and packed cotton are entirely exported.

80

contract negotiations, etc. The SAM includes three relevant MC accounts: MC for
domestic sales, MC for imports, and MC for exports. These accounts get payments from
the tradable commodity accounts (rows) and make payments to a trading services
commodity account (columns).

The Factor accounts describe the source of factor income, i.e., value added in
each domestic activity and from the rest of the world (RoW) and how factor payments are
distributed to the various institutions, including the different household groups and the
rest of the world in proportions reﬂecting factor endowments (column). The Zambezi
Valley SAM includes factor accounts for land (one), capital (ﬁve) and labor (four)
resources. The labor accounts correspond to two labor types (family and hired labor)
represented on a temporal dimension, i.e., each labor account is disaggregated to account
for labor use and remunerations for the pre-harvesting and the harvesting and post-
harvesting season, respectively. This allows for the analysis of seasonality in labor
demand and payments to institutions. The capital account is divided in four activity
speciﬁc accounts, each associated with a cash crop farm account, and one general capital

account associated all other activities.

81

023.5
3982 .80

van
Eat shaman.

30M

Lose _So

2: uo

mac—0:88;
ﬂaw—5.:

 

280 _So

mound“

mEocumzo;
2:85

W
.—.
a—
_

n.—
U
n.—
~—

cozuawcwh

5.552 w:=:=eoo< 38m Ruined FEE—Sm 8352—5 .3. 93:.

82

The Institutions accounts comprise all the income and expenditures of
households, government, and the rest of the world. The Household accounts record
income, i.e., the value of domestic factor incomes to households, inter-household
transfers, subsidies or transfer payments from Government, and remittances from the rest
of the world (row), and how households spend that income: household home consumed
outputs from the activities they engage in, consumption expenditures of marketed goods
and services, transfers to other households, payment of taxes, private savings and
remittances to the rest of the world (column). The disaggregation of households in the
Regional SAM is aimed at representing the dominant production structure implied by the
classiﬁcation of activities. With that in mind, we disaggregated the households in four
major groups: (i) Tobacco growing households: farming households in tobacco areas that
engage in tobacco production under contract with MLT or Dimon; (ii) Cotton growing
households: farming households in cotton areas engaged in cotton production under
contract with Dunavant or C.N.A.; (iii) Non-tobacco growing households in tobacco
growing areas; (iv) Non-cotton growing households in cotton growing areas. Each of
these groups of households engages in a portfolio of activities including food crop
production and non-farm businesses, or some form of wage labor.

Typically, the Government accounts collect taxes on income from activities,
commodities, factors, and transfers from the rest of the world (row), and pay for
Government consumption of goods and services, transfers to households and to the rest of
the world (column). The role of Government in this regional model is limited. In reality,
the ﬂows of the Government account in a sub-national SAM should reﬂect the actions of

two actors — the local and the Central Government. The Central Government in this SAM

83

is located in the “Rest Of the World”. For Simplicity, we consider a single Government
account representing a regional Government. Income sources for the local Government
include the collection Of a ﬂat lump-sum tax per capita in the region and a substantial
inﬂow from the “Rest of the World” that includes the Central Government. In the base
year that the SAM represents, there are no activity or commodity taxes on domestic or
foreign trade. Local Government expenditures include transfers to households in the form
Of pensions, and expenses in health and education at the local level.

The Rest of the World (ROW) accounts record payments to the rest of the world
for imports Of goods and services, the use of foreign factors, and Government transfers
(row), and receipts from exports, factor payments to domestic factors, remittances to
households, transfers to Government and foreign savings (column). Like in the case Of
Government, the spatial deﬁnition Of the SAM implies the existence Of at least two ROW
institutional accounts; the “Rest Of the Country ROW” representing the rest of the
national territory, and the “Foreign ROW” that represents agents outside Mozambique.
The ZVR-SAM, however, includes only a single ROW account that covers all the areas
outside the region, domestic and foreign. The base year that the SAM represents does not
include any taxes or tariffs on trade between the region and the ROW.

The Savings-Investment account records the savings made by all the institutions
(row) and how they are spent in investment goods (column). A detailed list of accounts
and levels Of disaggregation for the Zambezi Valley Regional SAM are presented in
Tables 4.2 and 4.3. The SAM database is used to calibrate the ZVR-CGE model detailed

later in this section.

84

Table 4.2. Regional SAM Accounts: Activities, Commodities and Marketing Costs

 

Accounts
Activities

Description Of Individual Accounts
Agriculture and Livestock/Fishing:

Tobacco Growing Farms - MLT

Tobacco Growing Farms - DIMON

Cotton Growing Farms - C.N.A.
Cotton Growing Farms -
DUNAVANT

Non-Growers - Tobacco Areas
Non-Growers - Cotton Areas
Livestock

Fishing

Processing/manufgcturing:

Food processing

Beverage processing

Other Processing/Manufacturing

Marketing/Export Of Cotton and Tobacc_o:
MLT - Tobacco Marketing and Export
DIMON - Tobacco Marketing and Export
C.N.A - Cotton Marketing and Export
DUNAVANT - Cotton Marketing and
Export

T_ransportation and Services:
Trading Services
Government Services

Other Services

 

Commodities

Cash crops:
Raw Cotton — Dunavant

Raw Cotton — C.N.A.
Packed Cotton — Dunavant
Packed Cotton - C.N.A.
Raw Tobacco - MLT

Raw Tobacco — Dimon
Packed Tobacco - MLT
Packed Tobacco — Dimon
Other agricultural raw and
processed commodities:

Maize Grain and Rice

Maize Meal

Other F lours
Bread/Biscuits/Pasta

Beans and Groundnuts

Root Crops: Cassava/potatoes
Vegetables, Green Leaves and
Fruits

Coconuts

Animals and animal products:
Meat - Cow

Meat - Goat

Meat - Pork

Meat - Birds

Fish and Sea/River Foods
Milk and Eggs

Processed foods/beverages:
Cooking Oil

Sugar

Salt

. Tea/Coffee and spicies

Prepared Ready to Eat Foods
Alcoholic Beverages

Services:

Education

Health

Trading Services
Housing/Water/Electricity/Maintenance

Agricultural inputs:

Seeds - Other Products

Pesticides - Ceperrnetrin + Acephate
Fertilizers - NPK + CAN + UREA
Other inputs

Non-foods and oth_er commodities:
Firewood and Coal

Fuel - Vehicles/Equipment/Spares
Wood/Grass/Cane Products

Textiles, wearing apparels and footwear
Metal/blacksmithing Products

Kitchen Utensils and other Home Apparel
Soap and Hygiene Products

Imported tobacco

Other Commodities

 

Marketing
Costs

Domestic Sales
Imports
Exports

Source: Zambezi Valley Regional SAM

85

Table 4.3. Regional SAM Accounts: Factors, Institutions and Savings-Investment

 

Accounts Description and Level of Disaggregation

 

Land Land
Capital Capital (general)
Farm speciﬁc - MLT
Farm speciﬁc - Dimon
Farm speciﬁc - Dunavant
Farm speciﬁc — C .N.A.
Labor Family Labor -— Pre-harvesting
Family labor — Harvesting and Marketing
Hired Labor — Pre-harvesting
Hired Labor — Harvesting and Marketing
Institutions Government Local Government
Rest of the World Domestic and Foreign Rest of the World
Households Cotton areas: Non-grower households
Cotton areas: Grower households
Tobacco areas: Non-grower households
Tobacco areas: Grower households

Factors

 

 

 

 

 

 

Savings- Savings-Investment Savings-Investment
Investment

 

Source: Zambezi Valley Regional SAM

4.3.1.3. The SAM as a Conceptual Framework

The SAM framework can be used as a conceptual tOOl tO explore the impact Of
exogenous changes in variables such as exports from the region, government
expenditures and a variety of investments on the sociO-economic system and the resulting
structure Of production, factorial, and household income distributions. In this sense, the
SAM becomes the basis for simple multiplier analysis and calibration Of CGE models
(Pyatt and Round, 1979, Dervis et al., 1982). The ﬁrst question to address when using the
SAM as a conceptual framework is to which accounts are considered exogenous and
which are endogenous. Sadoulet and de Janvry (1995) deﬁne endogenous accounts as
those for which changes in the level of expenditure directly follow any changes in

. . o l .
income, whrle exogenous accounts are those for Wthl‘l we assume that expenditures are

86

set independently of income. It has been customary tO consider the government, the rest
Of the world and the savings-investment accounts as exogenous, and activities,
commodities, factors and households as endogenous.37 To illustrate how the SAM
approach is helpful in deriving the changes in income distribution and expenditure
patterns by sociO-economic groups resulting from a change in the structure of production
due to exogenous shocks, Figure 4.1 presents the relationship among the endogenous
SAM accounts. These transformations incorporate the mechanisms that translate value
added from production into incomes of different types of households, via ownership of
factors Of production.

There are three key transformations tO consider. First, in addition to the
intermediate demand Of inputs by the activities, transactions T2], the level and the
structure Of output by different activities generate the aggregate demand for factors
(labor, land, and capital), which brings employment linkages into the analysis. The
stream Of value added from the production side rewards the factors Of production, with

wages going tO different types Of labor, and rents going to land and capital. This is

depicted in transactions T3.

 

37 An endogenous capital account would reﬂect some kind of internal ﬂexibility of capital ﬂows, while an

endogenous rest of the world assumes that trade is relatively free between the region and outside areas.

87

Figure 4.1

Relationships Among Endogenous Accounts

Intermediate Demand

 

 

 

 

 

 

 

 

 

 

    
 

 

 

 

 

 

A T21
T Production
Commodities Sales of Output Activities
l T12
Consumption
Patterns of Value added
Household Home by sectors
Groups Consumption T3,
T24 T14 ‘
Household Groups ‘ Factor Incomes:
» lntra-household Transfers wages and rents
T44
Factor Incomes

 

 

 

 

 

 

to Households
Tts

Note: The direction of the arrows indicates payments between the accounts.
Legend (subscripts i or j): 1 = activities; 2 = commodities; 3 = factors of production; and 4 = households.

Source: Adapted from Thorbecke (1998)

Second, the transformation from factorial distribution to the distribution of
income across household groups (transaction T43) depends on which groups own which
factors. For example, wage payment to hired labor goes tO households that provide this
type Of labor; that can be both in agricultural or non-agricultural sectors. Likewise, rental
income accrues to the owners of capital, land and other natural resources. Family labor is
rewarded at shadow wages. Resource endowment, particularly land and human capital, is
important at several levels. Landless households can be affected quite differently by

development policy than smallholders and large farmers (Thorbecke, 1998).38 Likewise,

 

3‘ Although landlessness, as such, does not appear to be a serious problem in Mozambique, there is a strong
positive relationship between access to land and smallholder welfare.

88

the better educated tend to be able to get jobs in better rewarded, more formal and
organized sectors, while the less educated are limited to employment Opportunities
largely in traditional agriculture and manual jobs Off-farm. Nevertheless, given the
current level of technological development in Mozambique, even the effects Of
employment linkages in those low technology sectors are likely to be important, if policy
is successful in generating increased rural output growth. sIt is worth noting that in the
study region part Of the hired labor force originates from Malawi, resulting in part of the
labor payments leaking outside the region. The design of the SAM and the CGE model
attempts to account for that.

The third transformation in Figure 4.1 (transaction T24) yields the consumption
patterns Of the different household groups as they Spend their additional incomes. It
reveals the value Of commodities/services purchased and consumed by those groups,
providing crucial information on their living standards, and providing indications Of
likely induced effects and the potential to strengthen growth through consumption
linkages, especially when they spend increasing shares Of their incomes on locally
produced goods.

Once the SAM is balanced, with all the analytically relevant disaggregation Of
selected endogenous accounts, it can be used to simulate the effects on the incomes Of the
endogenous accounts Of shocks in the exogenous accounts; government, savings-
investment or rest Of the world. This can be done using either ﬁxed price models
(multiplier analysis) or CGE models; the latter allow for shocks in model parameters,
such as tax rates, prices, sector productivities, and others. Appendix A details the SAM

multiplier model.

89

The SAM multiplier model is limited in several ways. First, it assumes that all the
endogenous sectors have unlimited capacity to supply goods and services, i.e., an
inﬁnitely elastic supply; a Keynesian demand-driven system without resource constrains
(Sadoulet and de Janvry, 1995, Taylor and Adelman, 1996).39 Second, prices are assumed
ﬁxed. This may be a reasonable assumption when the costs of trade with outside markets
are low and the region is likely to be a price taker for most goods and factors. However,
high costs for participation in those outside markets and imperfect substitutability
between family and hired factors (e.g., labor) may results in endogenous prices for a
number Of regional goods and factors in Mozambique. Third, SAM multiplier models
assume that production uses ﬁxed proportions technologies (constant marginal
productivity of factor inputs) and that average and marginal expenditure propensities are
the same. Relaxing the assumption of linearity in production requires the use Of CGE

models, which we discuss in the next section.

4.3.2. The Zambezi Valley Regional Computable General Equilibrium Model

4.3.2.1. Introduction

The Zambezi Valley Regional Computable General Equilibrium model (ZVR-
CGE) used in this essay is based on the IFPRI standard model Of Lofgen et al. (2002).
The Standard CGE model accounts for all the payments recorded in the SAM. The
calibration Of the model, therefore, follows the SAM disaggregation of activities,
commodities, factors and households previously described. All the relevant features

introduced in the SAM to reﬂect the local economy and the issues at hand are mirrored in

 

39 Appendix A presents an idea about some modiﬁcations to the model that allow to deal with this
limitation.

90

the CGE.

The model is written as a set of Simultaneous equations, most Of which are non-
linear. First order Optimality conditions capture the behavior of producers and consumers
assumed to maximize proﬁts and utility, respectively. Other equations include a set of
constraints that have to be satisﬁed by the system as a whole, but are not necessarily
considered by any individual actor. Such constraints cover markets for products and
factors, and macro aggregates, i.e., savings-investment balance, government budget, and
the balance of the current account Of the rest Of the world.

Figure 4.2 presents a stylized Structure Of the model, indicating the ﬂows between
the various SAM accounts. In this section, we outline the Standard model structure, and
add some considerations with respect to interlinked transactions and sector expansion,

and household typology and poverty analysis within the CGE framework.

Figure 4.2
Stylized Regional Model Structure

   

Factor
Markets

Domestic Private Savings

  
 

 

 
  

Factor
Costs

I Taxes Gov. Savings I

 

lnterrnediate

 

 

 

 

 

 

   
  

    

 

   

 

 

   
 

 

 

 

 

' Input Cost I I I [C l - lbs
I Actlvitles I___ Households Government ap tal: S l
, I Transfers
T I I
Ho e C t'
m onsump Ion \ Private . Government Investment
Consumption Consumption
Commodity . Demand
Sales Markets
Household Foreign
Remittances Transfers
Rest Of the I: Foreign Savings

 

 

 

 

World l

 

Source: Adapted from Lofgren et al. (2002)

91

The description of a CGE model structure is done by Specifying the agents and
their behavior, the rules that bring the different markets to equilibrium (Sadoulet and de
Janvry, 1995), and closures to system constraints. For the regional ZVR-CGE, we specify
(a) Activities, production and factor markets; (b) InstitutionszHousehOlds, Government
and the Rest of the World; (c) Commodity markets; (d) Macro closure rules for system
constraints; and (e) model calibration. This section draws mostly on Lofgren et al., 2002.
We focus on the essential elements Of the standard model and highlight features relevant
for the regional model. In Specifying the model structure, it iS important to keep in mind
the different sets and sub-sets Of the SAM accounts. Considerations on choice Of
functional forms and tradability Of commodities are sometimes differentiated across the

different account sets.

4.3.2.2. Activities, Production and Factor Market Closures

Each activity represents a producer that is assumed tO maximize proﬁts, i.e., the
difference between revenues earned and the cost Of intermediary inputs and payments to
factors. Proﬁts are maximized subject tO a production technology that follows the
structure represented in Figure 4.3. The top of the technology nest contains a Leontief
(LEO) function Of the quantities of aggregated factor value-added and aggregate
intermediate inputs.40 Value-added is deﬁned as a Constant Elasticity Of Substitution
(CES) of primary factors, and the aggregate intermediary input is a LEO function Of

disaggregated intermediate inputs that can be domestically produced or imported. Figure

 

‘0 We use the LEO alternative as the default for all activities. It should be noted, however, that a CBS
alternative at the top of the technology nest may be preferable for particular sectors if evidence supports the
idea that available techniques allow for the variation in the aggregate mix of value-added and intermediate
inputs.

92

4.4 illustrates the elasticity of substitution between factors and/or intermediates.

Figure 4.3

Structure of Production Technology

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

     
  

 

Commodity Outputs
(ﬁxed yield coefficients)
T
Activity Level
(CESILEO)
1

I 1
Value-added Intermediates
(CES function) (LEO function)

1

l 1 1 _ l
I commodities

 

I Imported ] LDomestic ]

 

 

 

[Hired 1 [ FamE] Wired?

Note: CES — Constant Elasticity of Substitution Function LEO — Leontief Function

Source: Adapted by author from Lofgren et al. (2002)

Each activity produces one or more commodities according to ﬁxed yield
coefﬁcients, e. g., a Tobacco Farming Activity produces tobacco, maize, and other food
crops. Likewise a commodity may be produced by more than one activity, e.g., tobacco is
produced by MLT and Dimon Tobacco Farms, and maize is produced by all agricultural
activities.“

AS proﬁt maximizing agents, producers choose their levels of production and
input use on the basis of prices in product and factor markets — each activity uses a set of

factors up to the point where the marginal value product of each factor is equal to its

 

4‘ Household survey data were used to compute the structure of crop production for the agricultural
activities. The data also provided the information needed for household non-farm activities represented in
the SAM. As explained earlier, the household groups deﬁned in the SAM were defrned on the basis of
information from the survey data and the structure of activities.

93

wage or rent. Factor wages/rents may differ across activities when markets are segmented

or even for mobile factors, when discrepancies emerge as a result of sector Speciﬁc

 

 

 

 

determinants.
Figure 4.4
Elasticities of Substitution
Factor or O = 0
Input 1
0 s o 51
o = inﬁnity
Factor or
Input 2
o = 0 Leontief: no substitution

0 s o 51 Cobb Douglas and CES: Imperfect substitution
0 = inﬁnity Perfect substitution

Source: Author

The Standard model offers alternative Factor Market Closures (FMC), i.e.,
mechanisms for equilibrating supply and demand in speciﬁc factor markets (land, labor,
or capital). A description of each available factor market closure follows and is

summarized in Table 4.4.

94

Table 4.4. Alternative Factor Market Closures

 

Factor Market Closures
Full

 

 

 

 

 

 

Variables Employment Unemployment gzaoremlﬁit
(Neoclassical) (FMOBUE) FESTFE
(FMOBFE) ( )
Variable Description F XD F LX FXD FLX F XD FLX
uanti Su lied of
01:80:) Factor? PP . . .
Quantity Demanded of
QFD (F’ A) Factor F by Activity A . . .
Economy-wide wage/rent
“3F (F) for Factor F . . .
WFDIST (F, A) Actrvrty-Specrﬁc Wage . .

Distortion for Factor F

Notes: FXD - Factor is ﬁxed; FLX —— Factor is ﬂexible; F — Factor; A — Activities. F ACTF E — Factor is
activity speciﬁc and fully employed; F MOBFE — Factor is mobile and fully employed; FMOBUE — Factor
is mobile and unemployed.

Source: Author

The ﬁrst factor closure is the Factor is Mobile and Fully Employed (F MOBF E).
The default closure is to ﬁx the supply of the factor at the observed base level, and allow
variation in an economy—wide factor price variable; e.g., land rental rate, wage rate, or
rate Of return to capital.42 This ensures that the sum of demands from all activities equals
the total quantity supplied in the system (full employment). Under this closure, faCtors
are mobile between the demanding activities. Each activity pays an activity-speciﬁc wage
that is the product between the endogenously determined economy-wide wage and an
exogenous activity-speciﬁc wage distortion term that is ﬁxed in this closure. This is also

called the neoclassical closure. Figure 4.5 illustrates this closure.

 

‘2 This economy-wide factor price variable is not speciﬁc to each activity. It is an aggregate variable used
to balance aggregate supply and aggregate demand of each factor. Each activity is assumed to have a price
distortion factor that will inﬂuence the activity speciﬁc factor price.

95

Wage/rent
rate
WF1

WFo

WFZ

   

 

Figure 4.5

Full Employment of F actors

Perfectly Inelastic Supply

QFSO

\

  \

 

QF2

QF1

QFO

 

QFS - Quantity of factor supply
QF — Quantity of factor demand
WF - Wage/rent (price of factor)

Source: Author

Quantity of Factor

A second closure rule is the Factor is Mobile and Unemployed (FMOBUE). This

assumes that a factor is unemployed and the real wage/rent is ﬁxed. In this closure, the

economy-wide variable is ﬁxed (exogenized) and the supply variable is endogenized.

Each activity is free to hire any desired quantity of the factor at its ﬁxed activity-speciﬁc

wage. In essence, the supply variable merely records the total quantity demanded. This

closure is graphically represented in Figure 4.6.

The Factor is Activity Specific and Fully Employed (FA C T F E) is a third closure

rule. Under this closure, the factor market is assumed to be segmented and each activity

is forced to employ the observed base year quantity, i.e., the factor is activity speciﬁc.

96

More generally, it is appropriate when there are Signiﬁcant quality differences (or activity
speciﬁcity) between units Of a factor used in different activities (Lofgren et al. , 2002).
In this closure, the quantity of activity-speciﬁc factor demands and the economy-wide

wage are ﬁxed while the activity speciﬁc wage terms and the supply variables are

 

ﬂexible.
Figure 4.6
Unemployment of Factors
Perfectly Elastic Supply
Wage/rent
rate

WF1 \

WFo . QFSo

WF2 \ QF,

QFO
QF2

 

 

 

QFs, QFSO ors1 Quantity of Factor

QFS — Quantity of factor supply
QF - Quantity of factor demand
WF — Wage/rent (price of factor)

Source: Author

4.3.2.3. Households, Government and Rest of the World

The institutional block of the standard regional CGE model describes the behavior

of household groups, Government, and the Rest of the WOrld. Households choose the

97

levels of conStunption that maximize their utility on the basis of disposable income and
prices. Household consumption covers marketed commodities, purchased at market
prices, and home consumed commodities valued at their Opportunity cost; the activity
Speciﬁc producer prices.43 This feature accounts for the simultaneous decision making
process of households as producers and consumers of certain commodities. That and the
use of activity speciﬁc producer prices for home commodities and market prices for
purchased commodities implicitly addresses the non-recursive nature of the household
decision making process in this model. Household consumption behavior of market and
home commodities is modeled according to Linear Expenditure System (LES) demand
functions, derived from maximization of a Stone-Geary utility function subject to a
consumption expenditure constraint.44

n
Max U = 2,6, ln(qc -q2) (4.1)

c=l
- " 0 " o
Sublect to Z chc + Z pc(qc - qc ) = E
c=l c=l
0

With qc ch
0<,Bc<l

Zﬂc=1

c=l

Where, qc is the quantity of consumption of commodity o, q? is the subsistence or

 

‘3 The standard SAM only disaggregates home consumption by activity and household, not by commodity,
activity, and household. In the regional SAM households consume from activities that produce multiple
outputs. To accommodate that, non-SAM data are needed to allocate home consumption across the
commodities produced by each relevant multiple-output activity. Shares of home consumption of
household farm outputs are derived from survey data for each agricultural, ﬁshing and livestock activity.

‘4 This utility function is a generalization of the Cobb-Douglas function and incorporates the idea that
certain minimum amounts of each good must be bought.

98

minimal amount of the consumption of commodity c that must be bought by the

household, ,BC is the marginal share of consumption of commodity c, pc is the price of

commodity c, and E is the total household consumption expenditure.
The ﬁrst-order condition results in a Linear Expenditure System (LES) which can
be written as:
pcqc = pcch + ﬂCIE - EJ119413] (4.2)
c:
This system can be interpreted as stating that expenditure on good c, given

as pcqc , can be decomposed in two components. The ﬁrst iS the expenditure on a “base

amount” q? of good c which is the minimum expenditure for which the consumer is

committed. The second is a fraction be of the supernumerary income, deﬁned as the

n
income above the “subsistence income”, ZquQ , needed to purchase base amount of all
Cd

goods (Intriligator at al., 1996). These two components correspond, respectively, to
committed and discretionary expenditure on commodity c.

Since household consumption for the various groups in the SAM (h) is for two
types Of commodities, from the market and from home production, we have a two-
component LES system, each with a structure similar to equation (4.2):

Consumption of marketed commodities

o 0 Oh Oh
1)qu = 19%,? + :iIEh - 2‘.quth - Z Zpactqacth (4.3)
c'eC aeA c'eC

99

Consumption of home production
h 0h m 0m_ Oh Oh
Pacqach= Pacqach + ach Eh " ZPC'q '2 ZPac'qac h (4-4)
c 'eC aeA c'eC

where,

E h household consumption expenditures

q C"; quantity of consumption of marketed commodity c for household h
q: C h quantity of home consumption of commodity c from activity a for
household h
p'cn market price of commodity c
h

Pac producer price of commodity c for activity a

,6"; subsistence consumption of marketed commodity c for household h

qa ch subsistence consumption of home commodity c from activity a for
household h

,BC"; marginal share of consumption spending on market commodity c for
household h

,8; h marginal share of consumption Spending on home commodity c from

activity a for household type h

Demand functions are derived by dividing both Sides of each equation by the
relevant price.

Government collects taxes and receives transfers from other institutions. The
standard model includes variables that account for direct taxes from institutions and
factors, value added and activities, import tariffs, export taxes, sales taxes, government
factor incomes, and transfers from the ROW. In our regional SAM, representing a base
situation and reﬂecting conditions on the ground, the regional government does not

collect taxes on activities or domestic and foreign trade. We only account for the

100

collection of IRN (Imposto de Reconstrucao Nacional) on a per head basis in rural
communities, and substantial ﬁnancing from the central Government and other
institutions in the Rest of the World.45 The government uses that income to buy goods
and services (e. g., health and education) and transfer to other institutions (pensions to
households). Govermnent consumption is ﬁxed in real terms and transfers are CPI-
indexed. Government savings (surplus/deﬁcit) is a ﬂexible residual in the model; more
details in Section 4.3.2.5 on macro closures.

The Rest of the World records transactions between domestic actors and the rest
of the world, including imports, exports, and transfers. Commodity trade with the rest of
the world is discussed in Section 4.3.2.4. Transfer payments between the rest of the world
and domestic institutions and factors are ﬁxed in foreign currency. Foreign savings, i.e.,
the current account deﬁcit, is the difference between foreign currency spending and

receipts.

4.3.2.4. Commodity Markets

All commodities, except home consumed output, enter markets. Figure 4.7 shows
the physical ﬂow for marketed commodities with the indication of quantity and price

variables relevant in each case.

 

‘5 In the policy simulations, however, the model allows us to introduce alternative ﬁscal and quantitative
trade policy measures.

101

Figure 4.7

Flows and Speciﬁcation of Aggregation
of Marketed Commodities

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Commod't
Output frclill Aggregate
Activity1 r-t (géplo'raté)
(QXAC | PXAC) Aggregate
ces CET
(...) Output
' QX PX
Commodity ( | ) Domestic Househmd
Oxthtuttfr om __, Sales Consumption
QXACMg)?AO (CD I PDS- (OH I PQ)
( I ) PDD) . +
065 Composrte
Commodityc— c(:50vernmtetr)lt
onsum l n
Aggregate (00 1 P0) (06 I :0)
Imports +
(QM I PM) Investment
(QINVlPQ)

 

 

 

Note: CES - Constant Elasticity of Substitution CET - Constant Elasticity of Transformation
Source: Lofgren et al. (2002)

For marketed output, the ﬁrst stage is the aggregation of each commodity
produced by different activities. As a result of differences in timing, location and quality
between different activities, these outputs are imperfectly substitutable. Therefore,

commodity aggregation is done using a Constant Elasticity of Substitution (CES)

 

function.
_ 1
ac ac _pac ac —l
QXC = ac 2506 QXACaC 6' Pc c 6 OX (4.5)
aeA
where
CX set of domestically produced commodities

102

QXC aggregate market production of commodity c
QXA C ac market output quantity of commodity c from activity a

age shift parameter for domestic commodity aggregation function
625 share parameter for domestic commodity aggregation function
pf.“ domestic commodity function exponent

Activity speciﬁc output is derived from the problem of minimizing the cost of
supplying a given quantity of aggregated output subject to equation (4.5). AS Shown in
equation (4.6), the optimal quantity of the commodity from each activity source (QXAC)
is found at the point where marginal cost of the commodity from that activity equals its
marginal revenue product. The activity speciﬁc quantity is inversely related to the

activity-speciﬁc price (PXAC).

—1
CC CC
PXA ca, = PXCQXCI ZagCCQXACgfc I ogg’QXAng’c ’1 a6 A, c e CX (4.6)
06/! i‘
where,
PXA C ac producer price of commodity c for activity a
PX C aggregate producer price for commodity c

The choice between commodities from different activities is cast as an
Optimization problem. Equation (4.6) is the ﬁrst-order condition determining proﬁt

maximization from selling the aggregate output, QXC , at the price, PX c , subject to the

aggregation function and disaggregated, activity speciﬁc, [commodity prices. For

commodities with a single producer, the value of the share parameter 6% is unity and as
a result QXACac = QXC and PXACaC = PXC.

This aggregated domestic output (QX) is then, in the second stage, allocated to

103

exports (QE) and domestic sales (QD). It is assumed that suppliers seek to maximize
sales revenue for any given aggregate output level, subject to imperfect transforrnability
between exports and domestic sales expressed by a Constant-Elasticity-of-
Transformation (CET) function.46 For all commodities with output, we have
PX CQX c = PDSCQDC + PECQEC c e CX (4.7)
For commodities with both domestic sales (set CD) and exports (set CE),47 such
as maize, groundnuts, potatoes, goats, etc, the CET function is given by

l

t t ‘7
QXC -_- a§[5§QEfC + (1 - of. page Ipc c 6 (CE 0 CD) (4.8)

F irst-order conditions for maximization of producer revenues given the two prices
(PDS and PE) subject to the CET function and ﬁxed quantity of domestic output QX,
indicates that the optimal mix between exports and domestic sales is deﬁned by the
Export-Domestic Supply Ratio

l

_—

QEC_ PEC 1—5; p54
QDc PDSc 5:.

 

c 6 (CE A CD) (4.9)

Equation 4.9 indicates that an increase in the export-domestic price ratio generates
an increase in the export-domestic supply ratio, i.e., a shift towards the destination that
offers the higher return.

For domestically sold output without exports and for exports without domestic

sales, the output transformation is given by

 

‘6 Following the small country assumption, export demands are inﬁnitely elastic at given world prices. The
supply price for domestic sales is equal to the price paid by domestic demanders.

‘7 The set of domestic commodities without domestic sales is referred to as CDN, and the set on non
exported commodities as CEN.

104

QXC = QDC + QEC c 6 (CD n CEN)U (CE n CDN) (4.10)

This equation allocates the entire output volume to one of these two destinations.
In the Zambezi Valley Regional Model there are a number of commodities that are
exclusively sold in the domestic market. For instance, all the raw tobacco and cotton
produced by farming households is sold domestically to Trading-Exporting Contract
Farming Firms that add value and export it packed to the rest of the world. The two
commodities are clearly differentiated. While the raw tobacco and cotton is all sold
domestically, the packed tobacco and cotton is all exported: there is no transforrnability
between domestic consumption and exports for these commodities. See Table 4.5, for
commodity tradability position in the Zambezi Valley economy.

Domestic sales (OD) and aggregate imports (QM) make up the composite supply
in domestic markets (QQ). Absorption, i.e., the total domestic spending on domestic
output and imports at domestic demander prices (net Of sales tax, but inclusive of cost of
trading inputs), is deﬁned as

PQCQQC = PDDCQDC + PMCQMC c c (CD U CM) (4.1 1)

In the standard model, the demand for these commodities is derived under the
assumption that domestic demanders minimize costs subject to imperfect substitutability,
captured by a CBS aggregation function, also referred to as the Armington Function.

1

q _ q
QQa =a§[6é’QM‘pc +(1-oglgoci’c] Pg ce(CDnCM) (4.12)

105

82.5888 850
82305.56 .8:an—
ocﬂw»: can meow
23:8: 8:85.
waaeaeeeﬁsoz
BoBSOto—cxo...
camohmﬁwhcooz’

3:..—

_e88;o use uooBBE
8.55-8? 850
mcoﬁztoﬁmoEocmom
mcoom

83:3: 93 $58:
8228 cher—r

£_8:\:o:8=cm
mowﬁo>on 0:382

Z<m 3530M xo=e> EBEeN ”825m

__o @380

mammazzzﬁ
8.5 l .82
ice I .8:
seem I .82
.38 l «82
$3580
33E
moESowo>
macaw—855820
mceom
muse—"582m
=85 8E2
82

53w BEE

8232 v8.8a

 

. £88 «8 3 bosom . 8258 33—
. ootouhﬂ. . 8:8 vac—8m
. :mﬂswsm . 8:8 252

ND w 0 EU w 6 Q0 w 0 3288383 MU w o )5 w O Q0 w o
58 com 38 am 88 com 3 com 58 com :28 com 88 com as com
mtoaxm 3895 83m 058500 83:56:50 mtomxm muons: 83m 38:59 8285800

 

moEcoEEoU 8.595

 

8588500 @88me

 

5:; seesaw 2.. a 8:85:80 2% a Essen; .3. see...

106

The domain of the CES function is limited to commodities that are both imported

and domestically produced. The optimal mix between imports and domestic output is

 

 

deﬁnedby
1
gLC: PDDC. 5‘? “Pg ce(CDnCM) (4.13)
QDC PMC 1-52.I

Equations 4.11 through 4.13 deﬁne the ﬁrst-order conditions for cost
minimization given the two prices (PDD and PM) and subject to the Armington function
and a ﬁxed quantity of the composite commodity (QQ). Equation 4.13 ensures that an
increase in the domestic-import price ratio causes an increase in the import-domestic
demand ratio, i.e., a shift away from the relatively more expensive source.

The composite commodity (QQ) is demanded in the domestic market in the form
of Household consumption (QH); Government Consumption (QG); Investment (QINV);

Intermediary input use (QINT); and demand for transaction inputs (QT):

QQc = Z QINTCa + ZQHct, + QG, +QINV, + QT, c e C (4.14)
aeA heH

4.3.2.5. Macro System Closures

The model allows for alternative closure rules for three macro balances: the
current government account (GOV-B), the savings-investment balance (SI-B), and the
current account of the balance of payments that includes the trade balance with the rest of
the world (ROW-B). Table 4.6 presents the alternative closure rules for each of the macro
balances and indicates default and alternatives chosen for the regional economy model.

The default closures in the regional model are chosen to best resemble the

circumstances in the regional economy. The closure for the Government balance assumes

107

that all taxes, and the real government consumption are ﬁxed and the Government
savings (the difference between its current revenues and expenditures) is a ﬂexible
(endogenous) residual. Although it is an available Option, given the limited role and
discretion of the local Government in this model, we do not consider alternative closures
where tax rates and government real consmnption are endogenously determined.

For the external balance, the standard model applied to national economies
normally assumes ﬁxed foreign savings with ﬂexible exchange rate. This particular
closure implies that, if, ceteris paribus, foreign savings are below the exogenous level, a
depreciation Of the real exchange rate would correct the imbalance by reducing spending
on imports and increasing earnings from exports. In sub-national models, the treatment
given to the ROW-B depends on the structure of “foreign” trade and ﬂows. For regional
economies that trade exclusively with the rest of the country (domestic ROW), the
exchange rate is assumed ﬁxed at the one-to-one rate (ﬁxed real exchange rate indexed to
the model numeraire) and foreign savings (and the trade balance) are allowed to vary. In
our model, however, given the evidence of massive trade between the region and the
foreign ROW added to the massive use of foreign currency in those transactions, it is
reasonable to account for a ﬂexible exchange rate regime closure rule.48 We will,
therefore, use the default of a ﬂexible exchange rate and ﬁxed foreign savings, but also

test the alternative of ﬁxed exchange rate and ﬂexible foreign savings.

 

‘8 Given the extremely poor road and communications infra-structure in the region, most of the imports into
the region are originated from Malawi. For the same reason, many of the identiﬁed exports, including
maize, potatoes, tobacco, and others are directed into Malawi, and a great deal of imports are originated
from that country. Factor payments to the rest of the world are also predominantly to Malawian migrant
labor. '

108

Table 4.6. Macro System Closures in the ZVR-CGE Model

 

Macro Balances and Available Closures

Macro Balances

Deﬁnition of Available Closures

Default and Alternative
Closures in the Regional
Model

Default Alternative
Closure Closures

 

Government Balance
(GOV-B)

Savings-Investment
Balance (SI-B)

GOV savings are ﬂexible, direct tax rates are
ﬁxed

GOV savings are ﬁxed, uniform direct tax rate
point change for selected institutions

 

GOV savings are ﬁxed, scaled direct tax rate for
selected institutions

Investment Driven Savings (Savings adjust to
given level of Investment) - Uniform mps rate
point change for selected institutions

 

Investment Driven Savings (Savings adjust to
given level of lnvestrnent) - Scaled mps for
selected institutions

 

wSavings Driven Investment (Investment level

deﬁned as a function of existing savings in the
economy)

 

Investment is ﬁxed absorption share — uniform
mps rate point change

 

Investment is ﬁxed absorption - scaled mps

 

Rest of the World
Balance (ROW-B)

Flexible Exchange Rate and Fixed Foreign
Savings

 

Fixed Exchange Rate and Flexible Foreign
Savings

I7Alternative macro closures tested in our experiments.

Source: Author

For the savings-investment balance, closures can be either investment driven (the

value of savings adjusts to pre-deﬁned investment levels) or savings driven (the value of

investment Simply adjusts to existing savings). We assume a savings driven closure,

under which investment is determined from available savings from households and the

local government. This implies that investment is endogenous and self-ﬁnanced by the

region. This is a reasonable assumption, given the missing and incomplete credit markets

in rural Mozambique (Benﬁca, 2003). The alternative closure (investment driven savings)

implicitly assumes that the govermnent is capable Of implementing policies that generate

109

the necessary private savings to ﬁnance ﬁxed real investment levels. Given the limited
Government modeling sophistication in the model, and under the current circumstances

in the regional economy, this is not a realistic closure.

4.3.2.6. Model Benchmark Calibration

The benchmark of a CGE model is a solution of the model that replicates the
observed economic data for the base year compiled in the SAM. The SAM records in its
entries nominal values of transactions in the base year. The ﬁrst step is to choose
measurement units for all factors so that all factor prices are initially equal to one.
Likewise, one chooses measurement units for the domestic commodities, imports and
exports so that the prices of domestic goods and imports, the world price of exports and
the exchange rate are all set equal to one in the base year. This normalization allows all
initial quantities and prices to be computed, and parameters, including those for the
production and utility functions that are directly computed from the shares to be derived
(Sadoulet and de Janvry, 1995). The shares matrix provides the starting point for
estimating parameters of non-linear, neoclassical production functions, factor demand
functions and household expenditure systems in the CGE model. Results in Section 4.4
present a great deal of outputs that illustrate base year factor, intermediary inputs, and
market and household expenditure Shares derived from the SAM.

In addition to the shares provided by the SAM, the calibration of the CGE model
requires that one deﬁnes three sets of elasticities: Production, Trade, and consumption.

Table 4.7 lists and deﬁnes the different elasticities within each set.

110

Table 4.7. Elasticity Sets and Deﬁnitions

 

 

 

 

 

 

 

 

 

 

 

Elasticity Sets Elasticities Deﬁnition

Production wgiggly”_Avggr»e_gation Output aggregation elasticity for commodity c
F actor-Factor Elasticity of substitution between factors
Substitution
Factor-Intermediate Elasticity of substitution between aggregate factor
Substitution and intermediate

Trade Armington Elasticity Elasticity of substitution between imports and

domestic output in domestic demand

Constant Elasticity of Elasticity of transformation between exports and
Transformation (CET) domestic supplies in domestic marketed output
Elasticity

Household Frisch parameter for Elasticity of the marginal utility of income with

Consumption whousehold LES demand . respect to income
Expenditure Elasticity — Expenditure elasticity of market demand for
Market Demand commodity c by household h
Expenditure Elasticity — Expenditure elasticity of home demand from
Home Demand activity a for commodity c by household h
Commodity Value Share Value share for commodity c in home

of Home Consumption
Source: Author

consumption of household h from activity a

The actual elasticity parameters used in the model were obtained thorough

derivations from survey data and from the country data available from the GTAP (Global

Trade Analysis Project) database.

The CGE model is homogeneous in all prices. A numeraire is chosen by ﬁxing an

aggregate price equal to one. The weights Of such aggregate price can be the initial

values of production in each sector. Commodity prices, wages and exchange rates are,

therefore interpreted in real terms.

4.3.3. Interlinked Transactions and Expansion Paths,

All raw cotton and tobacco production generated by smallholder growers in the

Zambezi Valley Region originates from contract farming arrangements organized by four

agro-industrial ﬁrms. As described in Chapter Three, those ﬁrms are assigned to speciﬁc

geographical concession areas, where they provide inputs and extension assistance to

111

small farmers on credit, and are granted monopsony rights that entitle them for the
purchase of all the output at predetermined prices. Given the failure in both input and
output markets, the interlinked arrangements imply that ﬁrms have a monopoly in input
markets and a monopsony in product markets. Given the lack of a wide range of
alternative sources of cash income, this type of arrangement is extremely important for
rural households in the region.

In the process of expanding their contract farming Operations, ﬁrms - that have
limited resources — will choose between alternative paths to do so. Essentially, they have
the option to support (i) relatively less farmers each receiving substantial level of support
(larger loans) and, therefore, planting larger areas with the crop; or (ii) a relatively lager
number of farmers each receiving smaller loans and, therefore, planting relatively smaller
areas with the crop.49 While this is an important issue, the choices faced to decide which
path is more appropriate depend on many factors that fall outside the scope of this study.

In this essay, we will assume that the contract farming operations represented in
the region expand the size of their operations with the same factorial structure
represented in the base scenario. As described later in this essay, expansion is combined
with series of other shocks. We are ultimately interested in assessing the income and

poverty effects of those shocks on the different household groups.

 

49Looking at the lenders’ point of view, Gangopandhyay and Sengupta (1987) prove that a capital
constrained lender interested in maintaining the level of beneﬁts of a previously less constrained
environment will Sign a smaller number of contracts with loans equal to or larger than the ones provided
under a less constrained situation. It can be argued, however, that given the concavity of the farmers’
production functions, it may be better for the lender to Sign a greater number of contracts with smaller loans
provided to each farmer. This depends, obviously on the stage of production where farmers are operating.

112

4.3.4. Representative Households and Poverty Analysis in a CGE Framework

4.3.4.1. Introduction

In this Section, we discuss a framework for the analysis of the impact Of
economic shocks on poverty Of household groups in a computable general equilibrium
framework. The CGE model described in Section 4.3.2 incorporates information on how
representative household groups earn and Spend their incomes. That sort of information,
including the rules governing factor markets, and household heterogeneity with respect to
factor endowments, demographic composition, consumption patterns and market access,
is important to assess the impact of exogenous shocks. Lofgren et al. (2003) point out
three features necessary for such a framework. First, it must include Shocks that are of
interest regarding their potential differential impact on household groups. Second, it
should be able to capture the impact of shocks on the extended functional distribution of
incomes; the distribution of incomes across disaggregated factors that remunerate the
household groups on the basis of ownership. Finally, it must map from this extended
distribution across factors to household incomes with enough detail to generate
information about the Size distribution needed to capture poverty and inequality
measures.

Approaches to undertake poverty analysis in a general equilibrium framework can
be aggregated into two major categories: Micro-Simulation (MS) and Representative
Household (RH) approaches. Each category has many variants.50 The essence of the MS

approach is to model the behavior of the individual agents, households and/or ﬁrms,

 

5° It should be noted that no single approach dominates and the choice is predominantly dependent upon
informational demands and operational constraints that vary across applications.

113

using a micro database linked to the standard CGE modelgthrough an integrated
CGE/Micro-simulation model or, in a sequential fashion, with the CGE model feeding
the micro-simulation model with price, income and employment data. Under the RH
approach, a separate module generates, for each simulation, results for individual
household income/expenditure by drawing on (1) a distribution function with known
parameters and known representative household incomes; or (2) individual survey
observations scaled using Simulated changes in representative household income from the
CGE model post-simulation results. Then, the results are used to perform standard
poverty and inequality analysis. Such analysis may include, but is not limited to, poverty
measures such as the F-G-T Pa class of indicators (poverty headcount, depth, and severity
of poverty) and stochastic dominance analysis, and various inequality measures. In this
study, given the nature Of the available data and the implied Operational feasibility, we
choose to use the RH approach. We now detail this approach and the essential rationale

and deﬁnitions for the stochastic dominance analysis used in this study.

4.3.4.2. The Representative Household Approach

The CGE models have a ﬂexible number of representative households (RHS). Our
ZVR-CGE model divides the households in two groups in each area, based on their
diversiﬁcation into cash cropping under contract. The RH approach assumes that,
following an external shock in the economy, the intra-group distributions shift
proportionally with the change in mean income. This means that the variance of each
distribution is considered ﬁxed, and exogenous to the model; if a shock increases mean
income by 5, the income of each household within a group is raised by 6.

Previous literature reports cases of signiﬁcant changes in intra-group distributions

114

in Asia in the mid-19805 (Huppie and Ravallion, 1991) and following the ﬁnancial crisis
of the 19903. However, more recent evidence suggests that inequality increases as often
as it falls during Spells of growth in developing countries, and that neutrality is a
defensible ﬁrst-order approximation (Ravallion and Chen, 1997; Decaluwé et al. ,
1999).51 In the absence of compelling evidence in either direction in Mozambique, we
adopt this neutrality assumption in our analysis. Dervis, de Melo and Robinson (1982),
stress that the complete endogenization of intra-group distributions following shocks
remains one of the biggest modeling challenges in studying income distribution in a
general equilibrium context.

The procedure allows us to undertake a comparative analysis Of the poverty
income Situation pre- and post-simulation. Such analysis can be done using (1) the
Foster, Greer and Thorbecke (F-G-T) Pa class of decomposable poverty measures that
allow the measurement of the proportion of the poor in the population (poverty incidence
or head count ratio) and the depth and severity of poverty (poverty gap and squared
poverty gap); 0r (2) the graphical comparative illustration on pre- and post-simulation
cumulative distribution functions (CDFs) of income. Following Deaton (1997), we

choose the second approach.52

 

5 ' In a cross-country setting, Gugerty and Tirnmer (1999) found that whether inequality raises or falls
depends on the initial distribution of assets; the broader the initial distribution of assets the more pro-poor
are the effects of growth.

52 In a CDF, the vertical axis goes from 0 to 100 and the horizontal axis shows our income measure.
Suppose that we have a sample of 100 households (or people) ordered from poorest to richest. The CDF is
just the graph of the observation number (which corresponds to percentile in this case) and the income
measure. Under that approach, if a shock leads the entire CDF to shift to the right then the new economic
environment stochastically dominates the base.

115

4.3.4.3. The Stochastic Dominance Approach to Poverty Analysis

Computations from our household survey allow us to rank the various household
groups by mean income. However, it is important to test whether the ranking is robust to
the choice of the poverty line.53 This leads to a special type of robustness test, referred to
as stochastic dominance, which deals with the sensitivity of the ranking of income levels
between groups to the use of different poverty lines. The Simplest way to do this - for the
robustness of poverty comparisons based on the headcount index Of poverty - is to plot
the cumulative distribution of income for different household groups. One can then see
whether the curves intersect. If they do not intersect, then the group with the highest
curve is poorer than the other group. If they do intersect, then for all poverty lines below
intersection, one group is poorer and for all poverty lines above the intersection, the other
group is poorer.

Formally, instead of worrying about the proportion of the population with income
below y, we consider y as being continuously distributed in the population with a
Cumulative Distribution Function (CDF), H (y) Let’s consider that there are two
household groups, with distributions H 1 (y) and H 2 (y), and we want to investigate
whether we can determine that one group is poorer than the other. We consider that
distribution H1(y) ﬁrst-order stochastically dominates distribution H2 (y) if and only if,
for all monotonic non-decreasing functions tp(y)

J’n

yl¢(y)dH1(y) 2 ¢(y)dH2(y) (4.15)
2’0 0

y

 

’3 Note that, while we are not computing measures requesting a speciﬁc poverty line, the validation of the
approach still depends on the existence of a relevant range of poverty lines.

116

where the integral is taken over all relevant levels of income. Considering that (p( y) is a
valuation function and that monotonicity implies that more is better, distribution 1 can be
labeled as better as it has more y, and it stochastically dominates distribution 2. An
alternative equivalent way of putting condition (4.15) is that, for all y,

H20) 2 H10) (416)

so that the CDF of distribution 2 is always at least as large as that Of distribution 1.
Graphically, it is always to the left of distribution 1 that is, therefore, ranked behind.

In discussing poverty, we need to follow a more restricted form of stochastic

dominance in which inequality (4.16) holds only over a limited range 20 S y S 21 , where

the z are alternative poverty lines. If we have two distributions H1 and H 2 , representing
two different household groups in a region, and we wish to ﬁnd out which one shows
more poverty and the extent to which the comparison depends on the choice of poverty
line 2, then if, for all poverty lines 2

112(2) > H1(z) (4.17)
the poverty incidence will always be higher for the distribution H 2 when compared to
distribution H1. TO test the robustness of the result, we need to graph the distributions we

want to compare, and, if one lies above the other over the range of relevant poverty lines,
the choice of poverty line within that range will have no effect on the outcome (Deaton,
1997)}:4

Figures 4.8 and 4.9, show hypothetical cumulative distribution functions for two

household groups, H1 and H 2 , in two different regions. In Region I (Figure 4.8),

 

5‘ The poverty rankings Of the distributions will be robust to all choices of the poverty line if and only if
one distribution ﬁrst-order stochastically dominates the other.

117

distribution H 2 lies everywhere above distribution H1. In that case, we can conclude
that the poverty headcount will be higher in H 2 than in H1, regardless of where we draw

the poverty line 2. In contrast, in Region 2 (Figure 4.9), the distribution functions cross, at
2*, so we can only make such statements for ranges of poverty lines. In this case, poverty

will be higher in H1 for poverty lines less than 2* and higher in H 2 for poverty lines

greater than 2*, where 2* represents a poverty line of close to $500 per capita.

Figure 4.8

Distribution Functions of Income per capita, by Household Group

Illustrative Example - Region 1

 

 

 

 

(I)
E
o 00, ~
.5
(D
U)
3
O
.C
is— O q
o o
C
.9
8
5.3 V. -
ID
.2
<3
3
E “I -
3
o

o -4

I I I I f
O 500 1000 1500 2000
Household income per capita (USD)
—— Household_Group_H1 ----- Household_Group_H2

 

 

 

Source: Author’s hypothetical distributions

118

Figure 4.9

Distribution Functions of Income per capita, by Household Group

Illustrative Example Region 2

.8 1
1 1_

.6
J_

Cumulative fraction of households

 

 

 

I
0 500 1000 1500
Household income per capita (USD)

 

 

 

Household_Group_H1 ----- Household_Group_H2

 

Source: Author’s hypothetical distributions

This approach allows for the visual comparison of the impact of alternative
simulations relative to the base. It also allows us to depict how income differentiation
changes among household groups following shocks to the economy, by looking at pairs
of CDFs (e.g., for different household groups) at the base relative to post shock scenarios.

In Section 4.4, we use survey data to illustrate a proﬁle of poverty and inequality
for the base year, along with other SAM based statistics. In Section 4.5 we undertake,

with the use of stochastic dominance techniques, an analysis following a number of

119

economic shocks that result in changes in the income of RHS in the CGE model that are
then used to scale household income in the survey database through the RH approach.
Since we assume inequality neutrality in comparing pre- and post shock scenarios, no
comparative inequality analysis is performed. The CGE model simulations are
implemented using GAMS and the Poverty analysis using the DAD (Distributive

Analysis/Analyse Distributive) Software.

4.4. Representative Characteristics of the Regional Economy
4.4.1. Introduction

This section provides a snapshot Of the Zambezi Valley Regional Economy. It
combines data from the household survey and the regional SAM to derive descriptive
measures of the regional economy in the base year, including demographic and economic
structure; structure of production, use of factors and intermediary inputs; remuneration of
factor incomes to institutions and structure of household income and expenditure
patterns; structure of domestic supply and demand, and foreign trade; and a poverty and
inequality proﬁle of rural households in the study area. The descriptive statistics in this
section is intended to set the stage for a better understanding of the impacts of policies

and exogenous shocks presented in Section 4.5.

4.4.2. Demographics and Economic Structure

The total population in the area is estimated at about 980,000 people, i.e.,
approximately 170,000 rural households. Out of those, 27,000 are cotton growing
households, representing 159,000 people, 46,000 are tobacco growing households,

representing 277,000 people, 66,000 are non-cotton growing households in cotton areas

120

and 31,000 are non-tobacco growing households in tobacco areas, representing 370,000

and 174,000 people, respectively.

Table 4.8 indicates that tobacco growers represent approximately 60% of the

population in tobacco concession districts, but only about 28% of the total Zambezi

Valley population. The incidence of cotton growers, within the concession districts and

across the region, is much lower at 29% and 16%, respectively.

Table 4.8. Population Data for the Zambezi Valley Region

 

 

 

 

 

 

Population Groups Number of Number of Share in Area Share in Total
People Households Population 1%) Population (%)
Cotton Growing Areas 528,317 92,900 100.0 53.9
Non-Growers 369,523 65,986 71.0 37.7
.__._--.qr.9>zers __-..._._l.5_§.7_9.4 26,914 2.9.9 162
Tobacco Growing Areas 451,069 77,248 100.0 46.1
Non-Growers 173,871 31,048 40.2 17.8
Growers 277,198 46,200 59.8 28.3
Total 979,386 170,148 - 100.0

 

Source: Computed from the ZV SAM-2003M and Survey data.

An analysis of the structure of the regional GDP shows that total absorption
(private and government consumption plus domestic investment) represents
approximately 94% of GDP, with private consumption being by far the most important
component (Table 4.9). As previously mentioned, the role of the regional Government is
limited which is reﬂected here in a negligible share of Government in GDP. The Zambezi
Valley economy is relatively Open, with the sum Of imports and exports representing 40%
of GDP. Given its relative self-reliance in food production and its orientation toward cash

cropping exports, the trade surplus of 6.4% of GDP is not surprising. Potential exists for

further increase in that surplus.

121

Table 4.9. Structure of the Regional GDP

 

 

 

 

 

Structure of Value Share
Regina] GDP ($US) (%)
Total Absorption 194,891,546 93 .6
Private Consumption 180,316,323 86.6
Investment 1 1,868,395 5.7
Government Consumption W ._....._ 2,706,827 1 .3
Trade Balance 13,325,918 6.4
Exports 48,306,452 23.2
Imports (34,980,534) (16.8)
Gross Domestic Product 208,217,464 100.0
Total Population (persons) 979,386
GDP per capita 212.6

 

Note: The values are estimated backwards, starting with the per capita incomes computed from Survey
data, then expanding with current population ﬁgures to actual GDP that gets allocated according to shares
extracted directly from the SAM (last column).

Source: Computed from the ZVR-SAM 2003/4 and Survey data.

GDP per capita in the region is estimated at $ 213, varying between $109 and $125
among cotton non-growers and growers, and between $175 and $318 among tobacco
non-growers and growers. Details on demographic comparisons among those groups of
households were discussed in Chapter 3. Appendix Tables B5 and B6 present a
Regional Macro SAM (at the highest level of aggregation), and selected characteristics of

the Zambezi Valley households.

4.4.3. Structure of Production, Use of Factors and Intermediate Inputs

As one would anticipate, the Zambezi Valley economy is predominantly
agricultural. Over 60% of the domestic production and domestic value added originates
in agricultural activities. More than half of domestic employment also originates from
agriculture (Figure 4.10). Although they conﬁrm the dominance of agriculture, these

numbers also indicate that the economy has a great degree Of diversiﬁcation.

122

Figure 4.10

Sectoral Structure of Value Added,

Production and Employment

 

l00 T
80 +
g: 60 I
40i
J: ,
m I
20 I
0 L
Value Added Production Employment
III Agricultural E Non-Agricultural

Source: Zambezi Valley Regional SAM

Table 4.10 presents the Shares of factors and intermediates by activity. This kind
of information is crucial in the analysis of the prevailing production technology in the
model. The CGE model structure considers how these two components get substituted

or complemented, assuming Speciﬁc technological options. Here we just look at

composition in the base year.

123

Table 4.10. Aggregate Factor and Intermediate Input Shares by Activity

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

Activities Value Added Share lnterrnediate Input Share
(%) (%)
Tobacco Farms
MLT 87.5 7' 12.5
DIMON _. W _ W 86.2 13.8
Cotton Farms
C.N.A. 98.0 2.0
Dunavant .. 98.5 1 .5
Non-Cash Crop F m
Non-Tobacco 94. 1 5.9
Non-Cotton 99.3 0.7
Other Primaﬂ
Livestock 66.2 33.8
Manufpcturing
Foods 55.3 44.7
Beverages 77.3 22.7
Other . “w 1.-...-” 1 00.0 . - .. 0.0
Services
Trading 65.3 34.7
Government 100.0 0.0
Other “M -.. 7 l .3 28.7
Marketipg/Export of Cash Crops
MLT 2.3 97.7
DIMON 2.1 97.9
C.N.A. 4.2 95.8
Dunavant 3. 1 96.9
Total 76.0 24.0

 

Source: Computed from the ZVR SAM-2003/4 and Survey data.

The analysis indicates that overall, value added constitutes 76% and intermediate
inputs only 24%. There are important variations across major activities. First, all
activities, with exception of the marketing/exporting operations, are predominantly value
adding activities. Second, within agricultural activities, we note that given its relatively
greater intensity in the use of chemical inputs, tobacco farms Show relatively higher
Shares of intermediate input use, 13-l4%, against just 2% among cotton growers and
0.7% among cotton non-growers. As expected non-tobacco growers in tobacco areas
present higher Shares on intermediates than farmers in cotton areas. This results from the

important technological spillovers that are occurring in tobacco areas, in the form of

124

increased use of chemical inputs by non-growers. Finally, while ﬁshing is a sector that

uses exclusively factors such as labor and capital and no intermediates, livestock,

manufacturing and some types of services (excluding government services) have a more

balanced composition of shares between value added and intermediate inputs.

Looking at the distribution of value added, production and employment across

commodities further illustrates the dominance of agriculture, particularly of primary cash

cropping sectors, such as cotton, tobacco, and maize. While tobacco provides by far the

greatest contribution to value added, production and employment, cotton frequently is not

much more important than maize (a major export and food security crop in the region),

vegetables, livestock and ﬁsh (Figure 4.11).

 

 

Employment

Maize grain
Livestock

I Government Services
8 Firewood/charcoal

Figure 4.11
Shares of Value Added, Production and Employment,
by Commodity
45.0 -
40.0 i
8\: 35.0 -
.... 30.0 I
*3 25.0 1
I: 20.0
8 15.0 ~
:3 10.0 ~ $2
'5, 5.0 . 5‘???
0.0 4
Value Added Production
a Cotton I Tobacco
III Vegetables El Other Agricultural
I Fish 251 Foods and Beverages
Tradin services 5 Housing and utilitres
E Wood grass/cane I Manufacturing goods

Source: Zambezi Valley Regional SAM

125

8 Other commodities

Figure 4.12 shows how factors of production get allocated across competing
activities in the base year. It is evident that tobacco farms absorb a great deal of land and
wage labor resources. In proportional terms, use of land by cotton growers and non-
growing farms in both regions is very similar. Use of family resources in the pre-
harvesting season is very much in line with the relative sizes of the population engaged in
speciﬁc activities, i.e., the predominance of tobacco growers and cotton non-growers. In
the post-season, there is a clear shift away from agriculture, with sectors such as

livestock, ﬁshing, manufacturing and services very much dominating.

Figure 4.12

Factor Shares Across Sectors of Activity

 

 

l00% -
’3‘
2‘, 80% -
'8
3
5 60%
Fe,
8.; 40% «
§
i 20% ~
0% 4
Land Family Pre- Family Hired Pre- Hired Capital
Harvest Harvest/Post Harvest Harvest/Post
In Tobacco Pants 5 Cotton Farm 0 Non-Tobacco
B Non-Cotton B Uvestock E Fishing
0 Manufacturing D Services I Marketing/Boon Ccmps

Source: Zambezi Valley Regional SAM

126

Tobacco farms absorb a great deal of hired labor in both seasons. Note, however,
that in the post season, the importance of service activities grows signiﬁcantly as the
demand for this type of labor by cash cropping activities and exported oriented activities,
such as maize (among non-growers in tobacco areas) shrinks. Appendix Tables 8.7
through 39, present a more disaggregated picture of factor shares across and within

sectors of activity.

4.4.4. Factor Incomes to Institutions and Structure of Household Income

In the SAM/CGE model, factors (labor, land, and capital) are remunerated to the
agents that own them, in proportions that mirror the structure of ownership. In our model
of the Zambezi Valley region those agents include domestic households that reside
permanently in the region, and foreign households that temporarily migrate to work as
wage laborers. That migrant labor, which in some cases consists of experienced former
tobacco smallholders, has played an important role in the recent expansion in the tobacco
sector. In economy wide models, we are particularly concerned with the fact that
payments to foreign households are potentially spent outside the economy, in which case
it generates economic leakages, as opposed to economic linkages that typically are
implied when incomes are spent domestically in domestically traded goods and
services.55

Figure 4.13 shows the distribution of factor payments between domestic and

foreign households. It shows that the share of total factor payments to foreign households

 

5’ While the importance of the Malawian labor cannot be ignored, we should also avoid overstating it. In
fact, a great deal of the Mozambican labor, both family and hired, has been exposed to tobacco growing in
Malawi in the recent past. Also, in reality, due to close family ties, a great deal of the income can actually
be spent in Mozambique as many people, in spite of their nationality keep resident status in both sides of
the border.

127

is relatively small, just over 5%. The importance of foreign households is greater when
we evaluate their share in the payments to hired labor in the region: they account for
about 18% of the total wage labor bill. We ensure that, in modeling the factor payments
in the SAM/CGE model we include factor payments to the rest of the world that mimic

this situation to avoid overestimation of beneﬁts to domestic households.

Figure 4.13

Factor Payments to Domestic and Foreign Households

All Factor
Payments

   

 

Wage Labor llllllllllllllllllll
Payments
0.0 20.0 40.0 60.0 80.0 100.0
Percent
Domestic Households [Ill Foreigners

Source: Zambezi Valley Regional SAM

Wage labor is particularly important among non-tobacco growers in tobacco
growing areas, for whom, as an income source, it is almost as important as the return to
family labor. Figure 4.14 presents the structure of factor incomes (and net transfers) for

each one of the household groups in the region.

128

Figure 4.14
Shares of Factors and Transfers in Household Income

501

Shares (%)

 

 

Non-Cotton Cotton Households Non-Tobacco Tobacco Households
Households Households

Bland llllFamily IWage ICapital IGOVTramfers lZlRemittances

Source: Zambezi Valley Regional SAM

There are several ﬁndings worth pointing out. First, in all cases, family labor is
the single most important factor of production, only rivaled by land among tobacco
growing households and wage labor among non tobacco growers. Second, households
in the tobacco area are net senders of remittances to outside the region. Third, we were
able to document a number of cases where households get relatively sizable (for rural
standards) government transfers, such as pensions to war veterans. Finally, returns to
land are signiﬁcantly higher among cash cropping households. Detailed results are

presented in Appendix Tables 8.10 and 8.1 l.56

 

5° To contrast with this factor remunerations approach, we produced a set of survey based “structure of
household income” statistics based on the income approach. See Appendix Table B. 12.

129

4.4.5. Structure of Household Expenditure

Income earned by households in the region can be spent on consumption goods
and services, saved, or transferred to other institutions, e.g., inter-household transfers,
transfers to households outside the region or taxes to the government. In our model, we
did not account for intra-household transfers, and transfers to the rest of the world were
netted out and entered as net income to domestic households from the rest of the world.
Transfers between domestic households and the government were also netted out and
entered as net payments from government to households. As a result, household
expenditure, as represented in the SAM, was reduced to home consumption, market
purchases of goods and services, and private savings.

Figure 4.15, generated from the Macro SAM, presents the aggregate structure of
household expenditure in the region. This structure is very typical of a rural developing
economy, with the share of home consumption as the dominant category, and the
prevalence of low savings rates.

Table 4.11 presents a more disaggregated structure of household expenditures that
identiﬁes sub-categories of home consumption and purchased goods and services, and
breaks down the results by household group. These results were generated using both
survey data to disaggregate the shares in home consumption, and the cash share

expenditures retrieved from IAF to input on the different household groups.57

 

’7 The IAF expenditures were attributed by ranking household groups in both data sets, for the relevant
enumeration areas as coded in the INE Master Sampling Frame. Although we averaged middle quartile
cash expenditure shares from IAF and attributed them to the two middle income groups in our data set, the
shares in the table for those two groups differ because we compute them over total income that includes
home consumption.

130

Figure 4.15

Structure of Household Expenditure

10.7

I Home Consumption
[11] Market Purchases
El Savings

66.8

 

Source: Zambezi Valley Regional SAM

Table 4.11. Structure of Household Expenditures - percent

 

 

 

 

 

 

 

Household Groups
Expenditure Cotton Areas Households Tobacco Areas Households All Region
Category Non- Growers Non- Growers
Growers Growers
Home Consumption 73.4 64.5 67.0 63.7 66.8
Agricultural goods 59.8 58.0 56.9 58.4 58.3
Maize 31.3 33.4 28.8 28.4 30.0
Other 28.4 24.6 28.1 29.9 28.4
Animal/Animal prods 13.4 6.4 10.0 5.3 8.3
Foods and beverages 0.2 0.1 0.1 0.1 0.1
Marketed 18.3 25.8 22.6 ‘ 24.0 22.7
Goods/Services
Agricultural goods 1.7 4.6 4.0 4.3 3.7
Animal/Animal prods 1.1 4.0 3.5 3.2 2.9
Foods and beverages 1.5 3.6 3.1 3.7 3.1
Non-food items 12.9 12.4 10.9 12.4 12.2
Education 0.4 0.3 0.2 0.2 0.3
Health 0.7 0.9 0.8 0.3 0.6
Savings 8.2 9.8 10.5 12.3 10.6
Total 100.0 100.0 100.0 100.0 100.0

 

Source: IAF and Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004

131

Several results stand out. First, all household groups have a high home
consumption share, dominated by the retained value of agricultural production and, to a
less extent, livestock, particularly among non-growers. Second, while in cotton areas the
value of retained maize is more important than all other agricultural products, in tobacco
growing areas the situation appears to be more balanced with maize and other aggregated
agricultural products exhibiting the same shares. Third, irrespective of household type or
ranking, retained value of home made foods and beverages are extremely low. However,
when it comes to marketed commodities, grower households tend to spend more in those
items. Fourth, spending in animal/animal products is higher among cotton growers and all
households in tobacco growing areas. Finally, against expectations, expenditure shares in
goods with (typically) high income elasticity of demand, e. g. non-food items, health and
education do not show any major differences across groups in the region. There are two
possible explanations for this ﬁnding. First, it may be that, within the income ranges and
preferences observed locally, demand for these goods is not income elastic. Second, the
supply of public sector services such as health and education is relatively restricted (low '
density of schools and health posts), which by itself limits access; the absence of a
reliable transportation system further reduces accessibility for all, regardless of income.
In addition, while the cash cost may be low, the opportunity cost of sending a child to

school, rather than having them work on the farm or in a business, can be high.

4.4.6. Domestic Supply and Demand, and Foreign Trade

As discussed earlier in this paper, the regional economy produces goods and
services that are home consumed, sold domestically, and/or exported to outside the

region. Some of the exported goods include cotton, tobacco, maize grain, groundnuts,

132

roots, vegetables, goats, alcoholic beverages, etc. While a great deal of domestic demand
is for goods produced domestically, many consumption items, particularly non-primary,
are actually imported from other parts of the country, or even brought in from
neighboring countries. Examples of imported items include rice, salt, sugar, tea, seeds
and chemical inputs, fuel, etc. See Appendix Table B. 1 3 for a detailed list of imported
and exported commodities and their importance. The Zambezi Valley economy is quite
open to trade. In the analysis that follows, we will ﬁrst look at the structure of imports
and exports by sector. Then, we will look at the commodity structure of imports and
exports in the region. Finally, we evaluate the share of exports in marketed output and the
share of imports in domestic demand.

Figure 4.16 shows that the regional exports are predominantly agricultural. Only
6.6% of the exported value is in non-agricultural products, including small livestock
(goats and birds) and home produced traditional beverages. In contrast, 95% of the
imports consist of non-agricultural products, most of which are directly imported from
Malawi.

Tobacco is by far the most important exported product in the region, about 74%
of the exported value, followed by cotton with 10% and maize grain with 4% (Figures
4.17 and 4.18). Other commodities with at least 2% of the total export value include
groundnuts, vegetables and goats. The most important imports are chemical inputs
(fertilizers and pesticides) and seeds, about 31% of the total import value, followed by
fuel with 19%, textiles/footwear and hygiene products with 9% each. Commodities with
at least 2% import share include rice, beef, cooking oil, sugar and salt. Appendix Table

B13 presents the results in greater detail.

133

Shares (%)

Share in Total (%)

 

Figure 4.16

Sectoral Structure of Exports and Imports

120 J

 

Exports Imports
E Agricultural 1111 Non- Agricultural

Source: Zambezi Valley Regional SAM

Figure 4.17

Commodity Structure of Exports

 

Commodities
I Packed cotton I Packed tobacco B Maize grain 1D G'oundnuts
I Root crops Vegetables l Meat — goat I Meat — birds
E Alcoholic drinks Other commodities

Source: Zambezi Valley Regional SAM

134

Figure 4.18

Commodity Structure of Imports

990.
‘
bﬁhﬁh
0999..
309000
9999..
to...
cocoon
.%&&&&33
$3.33. .

Share in Total (%)

 

Commodities
I Rice I Meat — cow I Cooking oil
11] Sugar L3 Salt 3 Fertilizers/Pesticide/Seeds
I Fuel 13 Textile/footwear I Metal/blacksmithing
I Soap and hygiene E Other Agricultural Other Non-Agricultural

Source: Zambezi Valley Regional SAM

In the CGE model, the allocation of marketed output between domestic sales and
exports, and the domestic consumption between domestically produced goods and
imports are functions of relative prices. Figure 4.19 presents the share of exports in
output and of imports in domestic demand for the base year. Of the total marketed output,
about 32% is exported and the remainder is sold in the local market. Note that
commodities such as cotton and tobacco are entirely exported, and about 50% of the
marketed maize is exported. On average, only 5.3% of the non-agricultural sales are
exported, against about half of the agricultural sales. This relatively high share of
exported agricultural marketed output may seem large. However, this result is not
surprising. In addition to the fact that cash crops are entirely exported, most of the
agricultural food production is actually consumed on the farm and domestic markets for

those food crops are relatively thin.

135

Figure 4.19

Shares of Exports in Output and Imports in Domestic Demand
by Sector

 

 

 

 

 

 

I .
Exports in Output Imports in Domestic Demand
5 Agricultural Ill Non-Agricultural Total

Source: Zambezi Valley Regional SAM

Regarding the share of imports in domestic demand, we ﬁnd a completely
different picture. About 27% of domestic demand value is imported. Among agricultural
commodities demanded, only 5% of the value is imported, while for non-agricultural
commodities over 94% is imported. This is consistent with the prevailing regional
commodity composition of imports, that is predominantly non-agricultural, and exports,

that is predominantly agricultural.

4.4.7. Poverty and Inequality Profile

In this section, we present a proﬁle of household income poverty and inequality in
the Zambezi Valley cash cropping economies for the base year using data from the
household survey. We use cumulative distribution curves of household income per capita

to undertake stochastic dominance analysis of poverty, as described in Section 4.3.4.3.

136

The analysis of inequality uses coefﬁcients of variation, Gini concentration ratios, and
Lorenz curves.

Income per capita differs substantially in the region between cotton and tobacco
concession areas, with the latter exhibiting a clear advantage. Figure 4.20 illustrates the
density curves for household income per capita by concession area, including all
households irrespective of their cash crop growing status. Several points are worth
noting. First, as expected in a rural African economy, both curves are skewed to the right.
Second, tobacco areas present a wider dispersion of income.

As compared to cotton areas, tobacco areas appear to have a larger proportion of
negative incomes per capita, but also a much greater share at the higher end of the
income spectrum. Cotton incomes are much more concentrated around the lower end.
Stochastic dominance analysis for the two concession areas, using distribution curves in
Figure 4.21, shows that for any relevant poverty line, households in cotton areas are

poorer than their counterparts in tobacco growing areas.58

 

’8 Note that as per the deﬁnitions introduced earlier, there is no ﬁrst degree stochastic dominance over the
entire income range; at unreasonably low poverty lines, the curve for the cotton areas dominates the one for
the tobacco areas. This is somewhat expected, as tobacco growers with low yields will incur large losses
due to high input costs.

137

Figure 4.20

Density Curves of Income per capita, by Concession Area
Zambezi Valley - Mozambique, 2003/2004

 

3

egg
org.

me per capita)
E E

o OM40?

 

 

 

 

 

 

-500 -250 0 250 500 750 1000 1250 1500 1750 2000 2250
Household income per capita (U SD)

1 - - - Tobacco areas — Cotton areas

 

 

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004

This CDF plot also conﬁrms indications from the density curve that incomes per
capita in cotton areas are signiﬁcantly lower. Nearly all households in the cotton area
earn incomes per capita lower than $400 per capita. More importantly, about 60% earn
$100 or less, and 80% earn $200 or less. Top incomes in tobacco areas are close to
$2,000, while 60th and 80th percentile incomes are about double those in cotton areas
($200 and $400, respectively). Median per capita incomes in cotton areas are

approximately $88 against $184 in tobacco areas. Computed mean incomes per capita are

$120 and $280, respectively.

138

Figure 4.21

Distribution Curves of Income per capita, by Concession Area
Zambezi Valley - Mozambique, 2003/2004

Cumulative fraction of households

 

 

 

I7 ‘T

0 1000 1500 2000
Household income per capita (USD)

or

 

 

 

Cotton_Areas ----- Tobacco_Areas

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004

Differences in income from the cash crop contribute substantially to these overall
income differences: mean and median income from cotton is $94 and $77, respectively,
while the same ﬁgures for tobacco are $731 and $364. In Figures 4.22 and 4.23 we
compare the cumulative distributions for growers and non-growers in tobacco and cotton
concession areas, respectively. ,

In both areas, for poverty lines at or below median per capita incomes,
comparisons are inconclusive. Indeed, in tobacco growing areas (Figure 4.22), the CDFs
overlap at very low levels of less than $50. At levels $50-$150 the grower households

dominate the non-growers, but the situation is again unclear around poverty lines near

139

area‘s median. For poverty lines deﬁned above the median ($184 per capita), and more
clearly above the mean ($280 per capita), grower households clearly dominate non-

growers.

Figure 4.22 1

Distribution Curves of Income per capita, by Household Group

Tobacco Concession Areas

Cumulative fraction of households

 

 

 

I I I I

00 1000 1500 2000
Household income per capita (USD)

 

Tobacco_NonGrowers ----- Tobacco_G rowers

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004

In cotton growing areas (Figure 4.23), poverty differences between the two
groups are very unclear for all poverty lines deﬁne below‘mean per capita income. For
poverty lines greater than the mean, i.e., over $120, growers clearly dominate non-

growers, indicating that poverty incidence is greater among non-growers.

140

Figure 4.23

Distribution Curves of Income per capita, by Household Group
Cotton Concession Areas

.8
t

.6
L

Cumulative fraction of households

 

 

 

l I f
200 400 600
Household incomerper capita (USD)

04

 

Cotton_NonGrowers ----- Cotton_G rowers

 

 

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004

In Table 4.12 and Figure 4.24 we present measures of inequality in the base year.
By all standards, inequality appears to be more severe in tobacco concession areas when
compared to cotton areas. This is illustrated by the higher coefﬁcient of variation, 1.17
(tobacco areas) to 0.78 (cotton areas), higher Gini concentration ratio (0.54 and 0.40), and
the fact that Lorenz curves for households in tobacco areas lie everywhere outside the

Lorenz curves for households in cotton areas.

141

Table 4.12. Measures of Inequality

 

 

 

 

Household Groups in Coefﬁcients of Variation of Per capita Income Concentration
concession areas per capita income Ratios (Gini)
Tobacco Areas
1.06 0-52
Non-grower households (0.165)
1.19 0-56
Grower households (0.102)
1.17 0.54
All households (0, 1 ()4)
Cotton Areas
0.64 0-35
Non-grower households (0.068)
0.86 0.44
Grower households (0.064)
0.78 0.40
All households (0.060)
_ _ 1.26 0.53
All households in the Region (0101)

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004.

Comparing inequality among groups within the same area shows, ﬁrst, that in
both areas inequality is more severe among cash crop growers. Second, these differences
are more accentuated in cotton areas, where Gini ratios exhibit a difference of eleven
points as compared to only 4 points in tobacco growing areas. The coefﬁcients of
variation give the same indication.

A visual interpretation of the Lorenz curves (Figure 4.24) provides further
insights. It further clariﬁes that, in cotton areas, inequality is more accentuated among
grower households, as their curve lies everywhere below the curve for non-growers.
Forty ﬁve percent of the non-growers of cotton receive about 20% of the total income
received by that group, while among growers, 45% receive only 15% of the group’s total

income. This higher level of inequality among cotton growers persists at all percentiles.

142

Figure 4.24

Lorenz Curves for Household Income per capita, by Household Group

Zambezi Valley — Mozambique, 2003/2004

 

_l l ‘
am 0.05 0.10

—+——-+

0.15 010 0.25 030
Cumulative °/o of households

I A
”—V— V

 

 

 

035 030 0.35 030 035 also 035 0.30 03 0.20 025 0.00 005 1.00

 

L -— Non-growers (cotton)

----- Growers (cotton)

- - - Non-growers (tobacco)

— Growers (tobacco) l

 

Source: Zambezi Valley Cotton and Tobacco Concession Areas Study Survey, 2004

The picture is a bit different in tobacco areas. For instance, at population

cumulative percentiles up to 40%, there are seemingly no differences between the two

household groups. At that level, in each group 40% of the population receive just over

5% of the total income of the respective group, which indicates a very high level of

inequality. Inequality becomes more severe among growers as we move up the curve;

60% of non-growers receive 25% of that group’s total income while 60% of growers get

only 20%.

143

4.5. Policy Simulations with the Regional CGE Model

4.5.1. Introduction

To assess the impact of expansion of cash crOpping activities, and of alternative
policies in cash cropping sectors on household per capita incomes, we have to consider a
number of alternatives with respect to the availability of resources, and the extent of their
mobility and allocation across various economic activities. In the CGE model we
accomplish this by deﬁning the speciﬁc mechanisms that guide factor market adjustments
in the presence of exogenous shocks. Those alternative mechanisms are discussed in
some detail in Section 4.3.2.2.

In Mozambique in general, and in the Zambezi Valley economy in particular, the
issue of availability and mobility of resources is important for various reasons. Post war
growth in agriculture has been primarily due to some area expansion (signiﬁcant relative
to the base, though still limited relative to its potential full employment) and labor force
growth, with limited gains in crop productivity. In fact, a great deal of the growth in the
Zambezi Valley cash cropping economies was possible due to the available pool of
resources; these included land for scheme expansion, and labor, some of which was
drawn from Malawi and from less dynamic domestic areas with fewer attractive
employment opportunities.

There are some fundamental questions in this context. First, under what
endowment/mobility scenarios can the Zambezi Valley economy continue to grow? In
other words, can the simple injection of additional resources by the contract farming
companies, without continued growth in land and labor supply and/or gains in the

productivity of the existing resources, ensure growth in cash crop production that is

144

capable of generating broad based growth in the regional economy? Second, how do
changes in world market conditions, e.g., ﬂuctuations in import prices of inputs and
export prices of outputs, affect the regional economy? Finally, how do all these changes
compare to each other regarding their effects on household income levels? In this
analysis, we consider selected exogenous economic shocks. While we have identiﬁed a

large menu of interesting experiments, our analysis will be limited to the following

shocks:
a. An increase in capital endowments speciﬁc to cash cropping.59
b. Productivity gains in cash cropping sectors;
c. Changes in world market conditions for tradable goods. The simulated

shocks include:
i. An increase in import prices for intermediate inputs, e. g.,
pesticides, fertilizers and seeds;
ii. An increase in export prices for cash crops (cotton and tobacco),
and maize grain;
d. A Government trade policy with respect to cash cropping sectors, e.g.,
export taxes.
The choice of experiments was based on an assessment of their importance for the
current policy debate in Mozambique, ensuring a mix of exogenously determined

variables, such as world pricesf’O discretionary variables like export taxes, and variables

 

’9 This includes the implicit increase in activity speciﬁc capital, resulting from increased support by ﬁrms
to smallholders in terms of extension assistance and other support that increases their managerial ability. In
reality, it is in ﬁxed proportion to intermediate inputs in those activities, reﬂecting, therefore, a
proportionally similar increase in the supply of intermediates by ﬁrms to smallholders.

Note that world market conditions may actually change in an opposite direction. Therefore, one should
keep in mind that there can be a change in the direction of the effects shown here. For example, a drop in

145

that can be inﬂuence by private sector actions such as productivity and the level of capital
injections in cash cropping schemes. The assumption is that the Government may have
the ability to inﬂuence private sector decisions through incentive mechanisms.

Some current circumstances, related to the choice of shocks include. First, there
is still room for expansion of agriculture in Mozambique. Given the absence of a wide
range of cash generating opportunities in rural areas, and the availability of land and
labor resources, cash cropping is viewed as a potentially important sector where direct
and indirect effects can be maximized. In the previous Chapter, the analysis indicates that
proﬁts in both sectors can be increased with the expansion in land area, but total crop and
total household incomes grow signiﬁcantly only in tobacco growing areas. We further
investigate, in this section, the economy-wide effects of expansion in both sectors.
Second, several studies in the cotton sector have emphasized low prices and poor
productivity at the farm level as factors leading to the stagnation of cotton farmer
incomes in Mozambique (World Bank, 2005; Tschirley et al., 2005). 6' The analysis in
Chapter 3, documents the low proﬁtability of the crop relative to tobacco in the Zambezi
Valley region.62 We will assess the impact, on the different household groups, of
increased productivity and prices; Third, a proposal to impose an export tax on raw
tobacco, as a way of encouraging domestic processing, has been the object of great

controversy in recent years. Benﬁca et al. (2005) discuss the implications of that policy in

 

world cotton/tobacco/maize prices will hurt household incomes and a decrease in import prices for inputs
will be beneﬁcial.

6‘ In fact, Mozambique pays the lowest prices in the region; the 1998-2002 average producer prices were
$0.16 per kilogram, compared to $0.22 in Zambia and Tanzania and $0.25 in Zimbabwe (Poulton et al.,
2001). Likewise, yields are amongst the lowest in Africa; 0.51 tonnes per hectare in 2003/04, compared to
0.9 tonnes in Zimbabwe and over 1.0 tonne in West Africa (Lemaitre et al., 2001).

62 In the tobacco sector, with the exception of some indirect complains blamed on poor grading by
companies, prices (and productivity) have not been much of a problem,

146

a cost-beneﬁt framework, and conclude that the imposition of such a tax at this stage of
development of the tobacco sector is not appropriate. We pursue the analysis of that
policy within an economy-wide framework to better inform its possible implications to
rural poverty. Fourth, given the importance of imported chemicals in the package
supplied to cash crop growers (fertilizer for tobacco, and pesticides in both sectors), we
analyse the implications of changes in their world prices to poverty in the study area.
Finally, maize is an important crop both for food security and as a cash crop for many
farmers in the region; some of it is actually exported to neighboring countries or deﬁcit
areas inside Mozambique. We assess the impact of changes in maize export prices to
poverty reduction in the study area.

We conduct the analysis separately for each concession area. The justiﬁcation for
this separation is that there are no overlaps of ﬁrm concession areas, so that a household
does not have the choice of moving into a different cash crop without physically moving
from the area. Although migration occurs, we found no evidence that supports the idea
that households were moving as part of a strategy to engage in alternative cash crops.

The implementation of the analysis on a sub-regional basis implied that the
regional SAM had to be divided into two independent matrices with similar structures.
The task of creating separate matrices was facilitated by the structure of the aggregated
SAM that included highly disaggregated activity and household accounts mapped out on
the basis of the relevant cash crops and conveniently linked by distinct factor accounts.
Background data from the ﬁeld survey originally used to generate the regional SAM,
including its non-farm accounts, was also used in this separation. The SAMs were

balanced separately through the Cross-Entropy method described in Chapter two and

147

used to calibrate two CGE models and undertake policy simulations.

Initially, for each area, we examined the simulations under the alternative
scenarios, considering individual impacts (each simulation individually), and also
combined simulations, where “a” (the injection of activity speciﬁc capital in cash
cropping sectors) is implemented in conjunction with each of the other simulations. The
scenarios experimented included: (a) full employment with full mobility of all factors; (b)
full employment of capital with unemployment and full mobility of labor and land; and
(c) unemployment and full mobility of all factors, except activity speciﬁc contract
farming capital that is assumed fully employed at the post shock level. Note that the
unemployment closure (c) implies that supply of resources is not restricted and can be
brought into production to meet the demand from expanding activities. Results for the
three scenarios are presented in Tables 4.13 and 4.14 for tobacco and cotton areas,

respectively.

Table 4.13. Effects on Household Income of Alternative Simulations: Tobacco Areas

 

--- % Charges in Householdrlncome per capita «-

 

 

 

Simulations Full Employment Semi-Unemployment Unemployment
Non- Growers Non- Growers Non- Growers
growers growers growers

Individual Shocks:

C. Farming Capital 1.42 -2.08 2.99 6.15 17.84 18.11
Productivity 4.21 3.34 8.43 1 1.63 17.95 30.09
Export price — Tobacco 3.94 3.79 4.01 4.88 23.19 44.02
Export price - Maize 2.24 1.50 5.90 6.94 31.38 53.06
Import price — Inputs -0.71 -1.32 -1.88 -2.84 -4.50 -6.37
WM “MExport tax — Tobacco -0.74 -4.52 -3.82 -7.28 -6.25 -10.4
Combined Shocks:
Productivity 7.36 1.35 22.58 21.20 43.16 61.98
Export price - Tobacco 4.14 1.66 9.30 9.14 49.01 79.04
Export price — Maize 3.42 -0.62 12.15 8.78 47.60 67.54
Import price — Inputs 0.51 -3. 16 3.34 —0.77 8.67 5.60
Export tax - Tobacco 0.02 -6.17 -0.78 -6.54 6.07 0.71

 

Notes: The individual shocks are in the magnitude of 15% in each case. The combined shocks include a
15% expansion in contract farming capital with another 15% shock. The Simulations use a Flexible
Exchange Rate Closure.

Source: ZVR-CGE Model Simulations.

148

Table 4.14. Effects on Household Income of Alternative Simulations: Cotton Areas

 

--- % Changgin Household Income per capita --

 

 

 

Simulations Full Employment Semi-Unemployment Unemployment
Non- Growers Non- Growers Non- Growers
growers Jowers ﬂowers

Individual Shocks:

C. Farming Capital 1.03 -l .92 4.43 1.22 14.20 13.46
Productivity 1.02 2.14 6.23 9.93 14.48 23.10
Export price — Cotton 0.64 2.06 2.21 4.07 5.71 9.33

Export price — Maize 1.80 2.12 4.71 5.42 21.44 31.06
Import price - Inputs -0.11 -0.21 -0.32 -0.45 -l .17 -l .56
Export tax — Cotton -0.16 -l .61 -1.02 -2.52 -2.20 —3.98

Combined Shocks:

Productivity 1 .34 0.30 6.06 5.84 31 .20 40.46

Export price — Cotton 1.93 -0.25 7.81 5.70 20.54 23.71

Export price — Maize 2.84 0.06 3.38 1.30 39.83 49.94

Import price — Inputs 0.91 -2.10 4.00 0.75 12.88 11.72
— -3.31 2.35 4.86 11.39

 

Notes: The individual shocks are in the magnitude of 15% in each case. The combined shocks include a
15% expansion in contract farming capital with another 15% shock. The Simulations use a Flexible

Exchange Rate Closure.

Source: ZVR-CGE Model Simulations.

In both areas, we ﬁnd that under full employment of all factors, closure (a),

economic expansion is very limited. Even if we assume that the existing factors are 15%

more productive, the effects on household income are very limited, and we ﬁnd little

mobility of factors across activities. Likewise, the effects of changes in world market

conditions, although resulting in the expected direction, are small. Under closure (b) the

effects are more sizable. However, the full employment assumption of non-cash crop

capital is highly unrealistic. Indeed, the non-cash crop stock of capital is very much like a

complement of household labor supply in the region that is assume available/unemployed

in this closure. This leads to limitations in the use of the available land and labor in the

expansion process by the activities that receive capital injections, and also in the

adjustment of other activities. In closure (c), where all factors are assumed to be available

for use in the expansion process, we observe that household incomes are responsive to

the various shocks. This closure also shov'vs some important indications of indirect effects

149

of shocks in cash cropping sectors on non-grower households in the adjustment process
through the marketing mechanisms.

We argue that closure (c) is the most realistic assumption in the Zambezi Valley,
for three reasons. First, population pressure, land degradation, and the resulting low
returns to agriculture in Malawi put increasing pressure on scores of people to migrate as
laborers; cash cropping areas of Mozambique are one of the obvious destinations.
Second, there are still localities in the study area that are not part of the concession
system, leaving room for further expansion of land and labor. Third, an increasing
number of people from other parts of Mozambique are willing to migrate to production
areas if the returns are compensating. Urban unemployment and lack of economic
opportunities are very real in Central Mozambique. This is true for both cotton and
tobacco growing areas. Finally, within each area, the proportion of land area still
uncultivated is relatively high,63 which means that more land can be brought into
production if additional labor is available in the region and capital is made available by
proﬁt seeking outgrower companies.

In the analysis that follows, we assume that all factors are fully mobile and
available for use, following an. injection of activity speciﬁc capital combined with a series
of other shocks. Table 4.15 presents, for each concession area, the list of shocks and the

effects on household income per capita by household type."4

 

63 In 1995, the total arable land in Tete Province was estimated in 49,000 Square Kilometers (SKMs), of
which 16 SKMs were used for permanent agriculture, 16,724 SKMs for shifting agriculture, and the
reminder was grass land, wooded grassland and shrub.

6‘ These results are based on a ﬂexible exchange rate closure for the rest of the world. These results are
consistent with the literature for developing countries.

150

Table 4.15. Effects on Household Income of Alternative Simulations, assuming
unemployment and full mobility of factors

 

--- % Changes in Income per capita «-

 

 

 

 

Simulations Shock Household Types
(% Change) Non-grower Grower Households
Households

Tobacco Areas

Contract Farming Capital 15.0 17.84 18.1 1
+ Productivity 15.0 43.16 61.98
+ Cash Crop Export price 15.0 49.01 79.04
+ Export price - Maize 15.0 47.60 67.54
+ Import price — Inputs 15.0 8.67 5.60
+ Export tax 15.0 6.07 0.71

Cotton Areas

Contract Farming Capital 15.0 14.20 13.46
+ Productivity 15.0 31.20 40.46
+ Cash Crop Export price 15.0 20.54 23.71
+ Export price — Maize 15 .0 39.83 49.94
+ Import price — Inputs 15.0 12.88 1 1.72
+ Export tax 15.0 11.39 8.59

 

Source: Zambezi Valley CGE Model Simulation Results.

For simplicity, we set all shocks at 15% from the base values represented in the

base year SAM. In the analysis that follows, we use the stochastic dominance approach

described earlier whose outcomes mirror the results in the Table. We examine the

impacts of each shock on the incomes of the two household groups in each area and

examine the mechanisms through which they arise. While, the SAM base scenario

represents a base year for an economy in equilibrium, it should be clear that model results

are only representative of the direction in which a system will begin to change towards a

new equilibrium until some (different) shock sets it on still another path; the length of run

for the impacts to take effect is therefore undeﬁned. In addition to that, population is

assumed constant during the adjustment process.

151

4.5.2. Policy Simulations in Tobacco Concession Areas

Households in tobacco concession areas are very responsive to exogenous shocks.
A 15% increase in capital speciﬁc to tobacco leads to an increase of approximately 18%
in the incomes of both household groups. If that injection is accompanied by an increase
in the productivity of the resources used in tobacco growing activities, both groups
beneﬁt signiﬁcantly, with average incomes of growers increasing 62%, against 43%
among non-growers.65 Figures 4.25 through 4.27 present the CDFs for those simulations

Figure 4.25

Comparison of Poverty Effects of Simulations, Tobacco Areas
Expansion, Productivity and Export Price

 

Cumulative fraction of households

 

 

 

000 2000 3000 4000
Household income per capita (USD)

-— Base --------- Expansion_CFC
----------------- CFC_Productivity ——— CFC_ExportPrice

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

 

 

 

6’ Expansion with higher tobacco world prices results in considerably higher growth rates, 49% in per
capita incomes of non-growers and 79% in the incomes of tobacco growers.

152

for all households in tobacco areas, and for growers and non-growers, respectively, based

on results in Table 4.15 and averages across all households.

Figure 4.26

Comparison of Poverty Effects of Simulations, Tobacco Growers
Expansion, Productivity and Export Price

 

 

 

 

 

‘- J —--' “wan- —- ——
U)
2
2 0°. 4
(D
8
O
J:
*5 to. -
C
.9
U
5:“ st —
(D
.2
E
a N. —
D
o
O
l I I I l
0 000 2000 3000 4000
Household income per capita (USD)
—— Base --------- Expansion_CFC
----------------- CFC_Productivity ——— CFC_ExportPrice

 

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

153

Figure 4.27

Comparison of Poverty Effects of Simulations, Tobacco Non-growers
Expansion, Productivity and Export Price

 

 

 

 

 

.8 1??"
2 0°. —
8
3
2
*5 <0. -
C
.9
‘6
1:" a. —
.3
E
a N. —
3
o

o -4

0 500 1000 _ 1500 2000
Household income per capita (USD)
Base --------- Expansion_CFC
l ----------------- CFC_Productivity ——— CFC_ExportPrice

 

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

An evaluation of the mechanisms through which these effects take effect can be
better understood, by stepping back to: 1) the ﬂow diagram presented in Figure 4.1, that
represents the generation of value added by economic activities, its remuneration to
households based the structure of factor ownership in the base year and household
consumption patterns; and 2) the structure of the economy in the base year represented in
the SAM and illustrated in Section 4.4. In this Section, we use key results to illustrate the
mechanisms in tobacco growing areas.

Model results indicate that initial growth generated in the tobacco sector generates

154

growth in the level of economic activity across all sectors. Table 4.16 presents the
structure of the level of economic activity in the base, and the changes in economic

activity by simulation.

Table 4.16. Base Shares and Changes in the Level of Activity by Simulation,
Tobacco Areas

 

Changes in Economic Activity by Simulation

 

 

 

Base
Economic Activities Alctivrty . Expansion Expansion Expansion Expansion
evel Expansron / / E rt / l t w/ Ex ort
(%) Only “’ . . W ?‘P° w PP“ P
Productrvrty Price Price Tax
Tobacco Farms -— MLT 35.1 15.8 47.6 31.5 9.3 4.1
Tobacco Farms — DIMON 10.0 15.4 46.6 30.4 9.7 1.3
Non-Tobacco Farms 14.7 15.9 42.2 49.1 6.6 10.3
Livestock 5.1 16.7 47.5 50.4 7.4 7.6
Fishing 3.6 18.0 53.2 60.3 7.3 7.2
Food processing 1.5 17.7 52.6 57.4 7.0 6.9
Beverage processing 0.9 8.5 ' 26.8 27.6 9.3 10.0
Other Processing 1.9 18.1 55.2 ' 65.1 6.4 6.9
MLT—Marketing/Export 12.4 15.8 47.6 31.5 9.3 4.1
DIMON—Marketing/Export 4.8 15 .4 46.6 30.4 9.7 1.3
Trading Services 5.0 15.6 47.6 43.7 8.5 5.2
Government Services 1.4 18.1 54.7 66.0 6.9 7.2
Other Services 3.6 18.0 53.1 62.0 6.6 7.4
Total 100.0

 

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

Results in the Table show that while agricultural activities, particularly tobacco
growing, and its associated value adding, dominate in terms of the base structure, growth
in economic activity in other activities is also substantial; for example, expansion with
productivity result in about 47% grth in the level of economic activity in tobacco
activities, 42% in other agricultural (non-tobacco) activities, and of over 47% in almost
all non-farming activities, including processing and services.

The relatively higher level of economic activity results in greater remunerations
for grower and non-grower groups alike, via the increased demand of factors by the

activities. For example in the case of expansion with productivity increases, Table 4.17

155

shows that tobacco growing activities increase their use of factors by about 30%, while
other non-growing farming activities expand even further, about 42%.?6 Demand for
labor by non-agricultural activities also expands dramatically. For example livestock,
ﬁshing, food processing, other manufacturing and trade, increase their demand for labor
resources by an average of over 50%. These increases are a result of consumption
linkages in the economy, as households demand additional goods and services with the

increased incomes.

Table 4.17. Changes in Demand for Factors from Activities in Tobacco Areas

 

% Change in Demand for Factors by Simulation

 

Economic Activities

 

Expansion Expansion Expansion Expansion Expansion

Only With . wrth Export wrth Input wrth Export
Productrvrty Price Price Tax
Tobacco Farms — MLT 15.9 30.8 35.2 8.9 3.5
Tobacco Farms - DIMON 15.4 29.7 3326 9.3 0.6
Non-Tobacco Farms 15.9 42.2 49.1 6.6 10.3
Livestock 16.7 47.5 50.4 7.4 7.6
Fishing 18.0 53.2 60.3 7.3 7.2
Food processing 17.7 52.6 57.4 7.0 6.9
Beverage processing 8.5 26.8 27.6 9.3 10.0
Other Processing 18.1 55.2 65.1 6.4 6.9
MLT—Marketing/Export 15 .8 47.6 31.5 9.3 4.1
DIMON—Marketing/Export 15.4 46.6 30.4 9.7 1 .3
Trading Services 15.6 47.6 43.7 8.5 5.2
Government Services 18.1 54.7 66.0 7.0 7.2

Other Services 18.0 53.1 62.0 6.6 7.4
Source: Base year ZVR-SAM and ZVR-CGE Model Simulations '
The actual magnitude of these impacts depends on the structure of factor use by
economic activity, presented in Table 4.18 (that determines the actual change in the
quantity used by each activity). As expected, tobacco growing activities demand the

greatest share of all factors; over three quarters of the land, close to 50% of the pre-

 

6" Note that the increase in demand for factors between MLT and Dimon is approximately the same —
30.9% and 29.7%, respectively, but the MLT base scenario demand level is much higher.

156

harvesting family labor, a quarter of the harvesting and post-harvesting family labor, over
50% of the waged labor throughout the season and about 45% of the capital. Demand for
family and waged labor by non-agricultural activities is higher in the post harvesting

season, when labor is released from agricultural tasks.

Table 4.18. Base Factor Use Shares by Activity, Tobacco Areas

 

Use of Production Factors
(Allocation across Activities)

---- o ____
Economic Activities /° 0f T0131 Base Year

 

 

 

Land Family labor ﬂit? Wage Wage Capital
PreHarvest PreHarvest PostHarvest
PostHarvest
Tobacco Farms - MLT 60.9 36.1 17.8 44.3 46.5 35.8
Tobacco Farms - DIMON 15.4 12.7 6.2 13.2 8.8 10.2
Non-Tobacco Farms 23.7 18.3 10.1 27.5 12.5 17.2
Livestock 0.0 12.2 15.5 1.0 2.9 0.0
Fishing 0.0 7.6 12.4 2.0 3.7 10.0
Food processing 0.0 1.4 3.1 2.0 0.2 1.1
Beverage processing 0.0 2.1 4.7 0.0 0.1 1.1
Other Processing 0.0 4.3 9.3 0.2 1.0 4.8
MLT-Marketing/Export 0.0 0.0 0.8 0.4 1.3 1.7
DIMONMarketing/Export 0.0 0.0 0.8 0.2 0.7 0.5
Trading Services 0.0 1.3 3.1 4.1 10.3 15.1
Government Services 0.0 0.0 0.8 4.1 8.1 0.0
Other Services 0.0 4.0 15.5 1.0 3.7 2.4
Total 100.0 100.0 100.0 100.0 100.0 100.0

 

Source: ZVR-CGE Model Simulations.

With the expansion and productivity increase, for example, additional factor
incomes (Table 4.19), i.e., the value of the quantity of factors (land labor and capital) at
market clearing wages/rents, generated in this process are substantial; remuneration to
land increases in about 33%, while that for family labor increases in 42% on average
(3 9% for pre-harvesting labor and 44% for harvesting and post harvesting), and for wage
labor increases in 55% on average (52% for pre-harvesting labor and 58% for harvesting
and post harvesting). Returns to activity speciﬁc capital are signiﬁcantly high in the

presence of productivity increases and increases in export prices. While increases in the

157

remunerations to activity speciﬁc capital as a result of a 15% injection, all else constant,
are only 22%, a compounded shock with a simultaneous 15% gain in productivity
generates an average increase of 300%. Since this capital is speciﬁc to grower households
the direct implications are exactly on that group. The non-grower group also beneﬁts
substantially from this, but only indirectly through the remuneration to other factors. As
noted in earlier sections, agricultural and non-agricultural activities in those areas are
labor intensive, with a mix of family and wage labor. The factor income payments to
each group reﬂect the original factor endowments in the SAM, summarized in Table
4.19. To a great extent, this helps explaining why impacts to non-grower groups are also
substantial. In Table 4.19, we can see that relative to returns to family labor, wage labor
(the fastest growing factor income) is relatively more important in non-grower
households, about 35% of total income, against only 22% among growers (percentages
from base structure). In that sense, an increase in labor hiring by growers and expanding
non-farm activities gets predominantly remunerated to non-grower households implying
that the income of those households increases accordingly.

It is worth noting that any expansion with or without productivity gains or price
increases is very compensating for non-grower groups. For instance, results indicate that
a simple expansion of 15% in the tobacco sector generates a 17.8% increase in non-
grower incomes (all reﬂected as indirect effects), and only 18.1% among growers (almost
entirely from direct effects).

If expansion in cash cropping schemes is accompanied by negative market events
such as an increase in the price of inputs, or a government imposed export tax on tobacco,

grth of household incomes slows down signiﬁcantly.

158

223:5; .882 mooéi as 2<m-x>N .8» van 02:8

 

 

159

 

 

3x: 458
2 3. new at. 4.2 2: .936
E- S _ 8.8 .. 2% SN 8.2 ”5.8332885: .59: 883
S 2: 8.8 3“ can ZN 85832.28 .533 ~me
3 3 as. n.3, . we 3: msexszhccagz :83 3:5
3 3 a? 2m 3. a. a assuage; :83 be:
S. 32 Sm 22.. ”.2 2: ES
. 830530: ooo<mo._.-zoz
92: .738
98- m. _ m- 93;. 33 as E .936 858% £23
2 3 3m 9:. 4.2 3. .355
E H: 0.8 2: 3m 3 msaxegwéagz .88.. owes
S .2: 0.8 Zn 3m 92 3:35:85 :83 8...?
2 3 ea. 3;. 3. 2; 85.32385: 553 been
3 o.» ”.9. 3m 3. v. _ N 3533..-»... .593 2:5:
3 3 23 3a ”.2 Nd“ 23
8402320: 83mg
58,—. toaxm outn— 59: 8E toaxm 33:62.2; i=0
5:: coﬁcanxm :23 c2885 5:: :oﬁcaaxm 5.3 coﬁcmaxm coacmaxm ax;
0.53—Ham 3309— new mU_O:Om:oI
ommm

case—:85 .3 2283::83— 38mm 5 mowcnzo

 

noun—35w .3 .muezﬁonsﬁom .8.an 2: ﬂ mow—:30 as 83.5 8mm .36 03am.

In our simulations, an expansion that is accompanied by an increase in the price
of imports for production inputs reduces income growth from 18% (in the injection alone
simulation) to 8.7% among non-growers and only 5.6% among growers that get directly
hit. This is illustrated in Figures 4.28 through 4.30 by distribution curves that just slightly
dominate the base, which means a higher poverty incidence, at all the relevant poverty
lines, relative to other simulations. Figure 4.28 shows the situation for all households in
tobacco areas, and Figures 4.29 and 4.30 present effects for growers and non-growers,
respectively. The export tax has the most devastating effects among growers whose
incomes literally stagnate as a result. These results have serious policy implications that
are discussed in detail later in this chapter.

Figure 4.28

Comparison of Poverty Effects of Simulations, Tobacco Areas
Expansion, Import Price of Inputs and Export Tax

 

Cumulative fraction of households

 

 

 

I I I

2000 2500

O—l

 

1000 1500
Household income per capita (USD)
— Base --------- Expansion_CFC
----------------- CFC_InputPrice ——— CFC_ExportTax

 

 

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

160

.4 .6 .8
L 1 1

Cumulative fraction of households
.2
|

 

Figure 4.29

Comparison of Poverty Effects of Simulations, Tobacco Growers
Expansion, Import Price of Inputs and Export Tax

 

 

 

 

 

0 500 1000 1500 2000 2500
Household income per capita (USD)
— Base --------- Expansion_CFC
----------------- CFC_InputPrice ——— CFC_ExportTax

 

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

161

Figure 4.30

Comparison of Poverty Effects of Simulations, Tobacco Non-growers
Expansion, Import Price of Inputs and Export Tax

 

.6 .8
4 1

.4
L

Cumulative fraction of households
.2
l

 

 

 

 

o .1
0 500 1000 1500
Household income per capita (USD)
— Base --------- Expansion_CFC
----------------- CFC_InputPrice ——-— CFC_ExportTax

 

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

4.5.3. Policy Simulations in Cotton Concession Areas

Household income levels in cotton areas are relatively less responsive to
economic shocks than those in tobacco areas. A ﬁfteen percent injection of activity
speciﬁc capital results in increases in household incomes of close to 14% for both
growers and non-growers. This indicates that, without changes in cotton prices and the
levels of productivity, the simple expansion of cotton production schemes results in
almost equivalent expansion in per capita income of grower and non-grower households.

Note that while grower income growth is predominantly attributable to direct effects,

162

non-grower incomes grow as a result of indirect effects.67 If that expansion is
accompanied by a 15% gain in cotton productivity, income growth is substantially
higher; 31.2% among non-growers and 40.5% among cotton growers. An increase in the
cotton export price generates income increases in the order of 20.5% among non-growers
and only 23.7% among growers. Figures 4.31 through 4.33, present the comparison of
poverty impacts of those simulations using CDFs for all households and separately for
growers and non-growers in cotton areas. The CDFs are directly computed using the

results presented in Table 4.15 for cotton areas and averages across all households.

Figure 4.31

Comparison of Poverty Effects of Simulations, Cotton Areas
Expansion, Productivity and Export Price

Cumulative fraction of households

 

 

 

 

0 200 . 400 600 800
Household Income per capita (USD)
Base --------- Expansion_CFC
CFC_Productivity ——- CFC_ExportPrice

 

 

 

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

 

67 Increases in incomes are a result of a combination of changes (positive or negative) in quantities of
factors used, and also changes (positive or negative) in wages/rents of those factors resulting from the
adjustment process. Those incomes are remunerated to households in proportions corresponding to their
original factor endowments.

163

Cumulative fraction of households

 

Figure 4.32

Comparison of Poverty Effects of Simulations, Cotton Growers
Expansion, Productivity and Export Price

 

 

 

0 200 400 600 800
Household income per capita (USD)
Base --------- Expansion_CFC
----------------- CFC_Productivity -—— CFC_ExportPrice

 

 

 

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

164

Figure 4.33

Comparison of Poverty Effects of Simulations, Cotton Non-growers
Expansion, Productivity and Export Price

 

 

 

 

.t- 4 2/1 .........
m
2
8 0°. 4
8
3
O
.C
“5 <0. ~
C
.9
‘6
g V. -
an
.2
L5
as! -
3
o

o —l

0 100 _ 200 _ 3 400
Household income per capita (USD)
Base --------- Expansion_CFC
----------------- CFC_Productivity ——— CFC_ExportPrice

 

 

 

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

As expected, all these expansions generate additional demand for goods and
services proportional to the increases in income for each group, but with variation across
the different items reﬂecting household consumption demand patterns. Like in tobacco
areas, we observe an expansion in the level of economic activity in all sectors. For
example, in the case of expansion with productivity gains, direct increases in the level of
activity of production and value-adding activities in cotton (averaging 44%) are higher
than increases in the level of activity of other sectors such as non cash cropping

agriculture (32%), livestock, ﬁshing, food processing, manufacturing, and services, that

165

range from 34-37%. These impacts are detailed in Table 4.20, along with results for

other simulations. The Table also shows the base period shares in economic activity.

Table 4.20. Base Shares and Changes in the Level of Activity in Cotton Areas

 

Changes in Economic Activity by Simulation

 

 

 

Base
. . . . t'vi
Economic Act1v1t1es Alfelvely E . Expansion Expansion Expansion Expansion
(%) x8351“ w/ w/ Export w/ Input w/ Export
y Productivity Price Price Tax
Cotton Farms — C.N.A. 24.6 14.6 44.7 20.2 13.7 10.1
Cotton Farms — Dunavant 3.7 14.5 42.7 19.7 13.6 11.1
Non-Cotton Farms 33.9 14.7 32.4 21.3 13.2 12.8
Livestock 7.9 14.6 35.2 21.4 13.2 12.1
Fishing 5.6 14.5 35.7 22.1 13.0 11.8
Food processing 2.2 14.5 36.6 21.4 13.1 11.8
Beverage processing 1.4 16.0 34.1 15.5 15.4 14.4
Other Processing 2.7 14.4 36.4 22.2 12.9 1 1.6
C.N.A.—Marketing/Export 4.5 14.6 44.7 20.2 13.7 10.1
Dunavant/Marketing/Export 0.5 14.5 42.7 19.7 13 .6 l 1.]
Trading Services 5.0 14.9 37.9 20.1 13.7 11.9
Government Services 2.4 14.4 36.2 22.4 12.9 1 1.6
Other Services 5.4 14.5 35.6 22.1 12.9 11.8
Total 100.0

 

Source: Base year ZVR-SAM and ZVR-CGE Model Simulations

In spite of that, the quantity of factors demanded from the cash cropping activities
is weaker in percentage terms relative to the demand generated in other sectors. In fact,
Table 4.21 shows that the quantity demanded of factors by activities with the productivity
enhancing path results in an average increase in factor demand from cash cropping
activities of about 26%, (27% among C.N.A. farmers and 26% among Dunavant
farmers). Note that C.N.A. operations are signiﬁcantly larger in the base year which
results is a larger absorption of resources in absolute terms. Demand for factors among
non-cash cropping farming activities increases 32.4%, while among all non-farming
activities it grows 32%-3 8%. In contrast, an expansion without productivity gains but

with an increase in cotton export prices, only results in increases in factors demand of

166

20%-22%, across all economic activities. Table 4.22 shows how factors are allocated

across the different activities in the base period.

Table 4.21. Changes in Demand for Factors from Activities in Cotton Areas

 

% Change in Demand for Factors by Simulation

 

Economic Activities

 

Expansion Expansion Expansion Expansion Expansion

Only With . wrth Export wrth Input w1th Export
Productrvrty Price Price Tax
Cotton Farms — C.N.A. 14.6 27.4 20.8 13.5 9.8
Cotton Farms — Dunavant 14.5 25.7 20.4 13.4 10.8
Non-Tobacco Farms 14.7 32.4 21.3 13.2 12.8
Livestock 14.6 35.2 21.4 13.2 12.1
Fishing 14.5 35.7 22.1 13.0 11.8
Food processing 14.6 36.6 21.4 13.1 11.8
Beverage processing 16.0 34.1 15.5 15.4 14.4
Other Processing 14.4 36.4 22.2 12.9 11.6
C.N.A.—Marketing/Export 14.6 44.7 20.2 13.7 10.1
Dunavant/Marketing/Export 14.5 42.7 19.7 13 .6 l 1.1
Trading Services 14.9 37.9 20. 1' 13.7 11.9
Government Services 14.4 36.2 22.4 12.9 1 1.6
Other Services 14.5 35.6 22.1 12.9 1 1.8

 

Source: Base year ZVR-SAM and ZVR-CGE Model Simulations

Table 4.22. Base Factor Use Shares by Activity, Cotton Areas

 

Use of Production Factors
(Allocation across Activities)
--- % of Total Base Year ---

 

Economic Activities

 

 

Land Family Famrly Wage Wage .
labor labor PreHarvest PostHarvest Cap Ital
PreHarvest PostHarvest
Cotton Farms — C.N.A. 37.9 23.1 23.8 13.4 46.6 20.4
Cotton Farms — Dunavant 4.9 4.8 2.0 4.4 3.9 4.2
Non-Cotton Farms 57.2 41.1 19.0 41.5 15.8 44.9
Livestock 0.0 11.3 12.9 3.9 3.3 0.0
Fishing 0.0 7.4 11.5 6.6 5.0 10.7
Food processing 0.0 1.1 2.3 4.0 0.3 1.0
Beverage processing 0.0 2.0 3.7 0.0 0.1 0.7
Other Processing 0.0 4.1 7.9 0.7 1.3 4.8
C.N.A.—Marketing/Export 0.0 0.0 0.0 0.3 0.8 0.8
Dunavant/Marketing/Export 0.0 0.0 0.0 0.0 0. 1 0.]
Trading Services 0.0 1.1 2.6 9.1 8.6 10.0
Government Services 0.0 0.0 0.0 13.1 9.9 0.0
Other Services 0.0 3.9 14.3 2.9 4.4 2.4
Total 100.0 100.0 100.0 100.0 100.0 100.0

 

Source: ZVR-CGE Model Simulations.

167

The resulting changes in factor remunerations (additional value of factor
payments at market wage/rents), under the productivity gain path, are fairly balanced
reﬂecting variations in wages/rents; remuneration to land increases in about 30%, while
that for family labor increases in 33% on average (31% for pre-harvesting labor and 33%
for harvesting and post harvesting), and for wage labor increases in 32% on average
(33% for pre-harvesting labor and 31% for harvesting and post harvesting). Returns to
activity speciﬁc capital average 200%, reverting exclusively to grower groups (see Table
4.23). While land is evenly distributed, non-growers, as the most populous group, have
most of the labor available in the economy, and beneﬁt from wage labor, particularly in
the ﬁrst part of the season (pre-harvesting) when it supplies over half of the wage labor.
In the post harvesting, the initial employment between the two groups is quite similar
with each group supplying about half of the labor.

This is a clear indication that productivity increases in cotton production allow for
good prospects for expansion not only in the sector itself, but also in other sectors of the
economy where resources can be productively employed, which leads to greater
possibilities for increase in household incomes irrespective of cash cropping status. As
pointed out in the beginning of this section, any expansion in cotton production results in
some indirect employment effects to non-growers, even when beneﬁts to growers are
limited.

A somewhat surprising result in cotton areas is that, although slowing growth, the
effects of adverse circumstances are less severe than in tobacco areas, although, as
indicated in the Figures for both groups, poverty is more severe as compared to other

scenarios. In fact, an increase in input import prices during expansion reduces income

168

growth from 14% to an average of 13% among non-growers and 12% among growers. It
is clear that that scenario still dominates the base, indicating that poverty reduction
occurs, but its magnitude is insigniﬁcant. A possible justiﬁcation for this is that the input
package in cotton areas is much cheaper than that in tobacco areas. On that basis, one can
infer that price increases of this magnitude are not enough to shake the costs of cotton
growers to the extent of impacting poverty signiﬁcantly. The export tax has also a
relatively small impact on poverty, but as expected grower households are relatively

more negatively impacted.

169

25.3335 .232 602% .55 2<m-...>~ so» 86 .858

 

 

 

 

ode Jar—10H
...N. N. N. N. . N N...N S: ...N. .336
o. . 2. .. . N N. . N 3.. ...m 3.952358%: 3...... ”we;
9: ~.m . Em . ..mm v.3 v.9 . 33832-0... .893 own?
6... mm . N. .N QNN 0.3 3: menacezhczmota: .836 3:3”.
a. .. To. . ...N a. . N N... 02. $5822-»... .836 2.5....
0... ma. ...N Ndm 53 «.3 , 2.3
mQAOIMmDOI ZOEOUiOZ
. odo— 12.—.0...
...N. a... on. ...OON no. to .536 2.38% 33.3.
...N. N.N. N. .N Ni 5.. 5 .536
c... on. .. .N N. _ N 3.. n: 3.35.32.33.33: 3...... owe?
a... NM . m. .m ..mm v.3 o. 5 353.365 .853 owe?
e... ...N. N. .N N..N... 3.. ...N. 35.323385: 3...... 3.3....
N... 2. ...N a. . m 5.. <2 m..__....oe.....e.. 3...... 2.3....
6... mm. ...N mdm N23 NKN vent—
moqozmmDOI 20:00
x3. toaxm 83m 59.— 3.5 tonxm b33059... i=0
.23 coacmaxm 5.3 coﬁcmaxm 5:5 coacaaxm 53> coucmgxm coicmaxm 3...
0.53—Cam Ego—wk van mU_o:om=o:
omwm

5.8385 .3 20:86::on .88“. E Swamp—U

 

note—55m .3 acetate—.53— ..Suah 2.. E mow—:30 can 8.3.5 omam .m~.v 035.

170

Cumulative distribution functions of the comparison of effects of the various
simulations are presented for all households in cotton areas (Figure 4.34), as well as for

growers (Figure 4.35) and non-growers (Figure 4.36).

Figure 4.34

Comparison of Poverty Effects of Simulations, Cotton Areas

Expansion, Import Price of Inputs and Export Tax

 

 

Cumulative fraction of households

 

 

 

I I

O 200 400 600

—

 

 

Household income per capita (USD)
Base --------- Expansion_CF C
---------------- CFC_InputPrice ------ CFC_ExportTax

 

 

 

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

171

.6 .8
L 1

Cumulative fraction of households
.4
1

 

Figure 4.35

Comparison of Poverty Effects of Simulations, Cotton Growers

Expansion, Import Price of Inputs and Export Tax

 

 

 

 

0 200 400 600
Household income per capita (USD)
Base --------- Expansion_CFC
----------------- CF C_lnputPrice ——— CF C_ExportTax

 

 

 

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

172

Figure 4.36

Comparison of Poverty Effects of Simulations, Cotton Non-growers

Expansion, Import Price of Inputs and Export Tax

Cumulative fraction of households

 

 

 

 

 

 

 

o —4
0 100 200 300
Household income per capita (USD)
Base --------- Expansion_CF C
----------------- CFC_InputPrice ——— CFC_ExportTax

Source: Base year Zambezi Valley Study Survey data ZVR-CGE Model Simulations

4.6. Summary of Policy Implications

A fundamental question in this essay is related to the extent to which strategies
that rely on indirect effects can effectively be used to target rural poverty. In other words,
can shocks to cash cropping sectors generate broad based income growth and poverty
reduction effects? If so, how strong are those indirect beneﬁts? We address this issue
through the analysis of the poverty impacts of selected policy simulations — favorable and
unfavorable shocks occurring concomitantly to expansion of contract farming schemes in

cash cropping economies in the Zambezi Valley of Mozambique. Such shocks include

173

technology shifts, trade issues including increases in prices of imported inputs, changes in
export prices of maize, cotton, and tobacco, and government export taxes.

In tobacco growing areas, expansion in the sector spreads growth to virtually all
sectors of activity, including non-cash cropping agriculture and non-farming activities
such as manufacturing and services; these growth results in greater remunerations for
grower and non-grower households, via the increased demand of factors by those
activities. The analysis suggests that in these areas, the beneﬁts of expansion with higher
tobacco prices have a very important poverty reduction impact, particularly among cash
crop growers, even when productivity is assumed unchanged. However, a more balanced
growth in incomes of growers and non-growers is obtained when productivity among
growers increases. Increases in import prices of inputs and Government imposed taxes on
tobacco exports signiﬁcantly slow down income growth in rural areas, particularly among
grower households. Since export price increases are uncertain,- and almost as likely to
occur as reductions, a productivity increasing effort is worth pursuing in tobacco areas to
compensate for any losses resulting from exogenous factors, including increases in
import prices of chemical inputs and seeds, and even reductions in export prices.
Associated interventions can include strengthening of extension systems to ensure the use
of proper techniques/ﬁeld practices, better grading, and improved environmental
management. Regarding foreign trade policy, the ﬁndings suggest that the Government
needs to avoid the use of trade restrictions, such as export taxes, as they may undermine
poverty reduction efforts by signiﬁcantly slowing down income growth. On the other
hand government and private businesses need to ﬁnd ways to minimize the importing

costs of intermediary inputs. While Mozambique has no power to inﬂuence world prices

174

of imported intermediaries, any policies that prevent further costs increases and actions
that can help reducing transaction costs are welcome.

In the cotton sector, several studies have emphasized low prices and poor
productivity at the farm level as factors leading to the stagnation farmer incomes in
Mozambique (World Bank, 2005; Tschirley et al., 2005). 68 This essay documents the low
proﬁtability of the crop relative to tobacco in the Zambezi Valley region. The concession
model as applied in Mozambique, which precludes competition among companies and
does not balance this with any effective performance monitoring system, must be
considered an important contributor to the problem of low prices and also low
productivity (Tschirley et al, 2006; Poulton et al. 2004); management subjected to little or
not competitive discipline will be able to transfer inefﬁciency costs to farmers through
low prices.

Simulation results indicate that, although limited when compared to tobacco
areas, expansion in the cotton sector, even when beneﬁts to growers are small, generates
some expansion in non-cotton sectors where resources are productively employed. When
that expansion is accompanied by productivity gains in cotton there is a much stronger
broad-based income growth and poverty reduction effect, even greater than increased
world prices. From a policy standpoint, because sustained increases in world prices are
unlikely, this result is encouraging. All this suggests that continued expansion efforts in
cotton focused on dealing with institutional issues and productivity enhancing

technologies, as they succeed, will increase those beneﬁts signiﬁcantly. Interventions to

 

6‘ In fact, Mozambique pays the lowest prices in the region; the 1998-2002 average producer prices were
$0.16 per kilogram, compared to $0.22 in Zambia and Tanzania and $0.25 in Zimbabwe (Poulton et al.,
2004). Likewise, yields are amongst the lowest in Africa; 0.51 tonnes per hectare in 2003/04, compared to
0.9 tonnes in Zimbabwe and over 1.0 tonne in West Africa (Lemaitre et al., 2001). Tschirley et al (2006)
assess the impacts of the sector’s regulatory structure on this performance.

175

improve the business environment for the emergence and sustained growth of non-farm
businesses are also important to fuel further growth.

Lastly, a set of additional policy considerations are worth pointing out. We ﬁnd
that, in both areas, households are heavily engaged in maize production and trade, and
can beneﬁt substantially from high export prices.69 Beneﬁts are more sizable among cash
crop growers but poverty reduction among non-growers is also remarkable. While high
export prices of maize are desirable, they are outside the control of Government and
individual producers and traders. The important policy implication here is that
Government needs to allow continued ﬂows of maize between locations inside
Mozambique, and exports to neighboring countries, such as Malawi and Zimbabwe,
when those countries are faced with cereal deﬁcits.7o Given the limited effective demand
inside Mozambique, and the high unit transportation costs in that domestic trade, this is a
good opportunity that can help reduce poverty in rural communities on a broad based
fashion, especially at times when prices are high. In this line, it is suggested that contract
farming schemes in both areas include improved maize seeds, and productivity enhancing
technologies to increase maize yields and maximize these potential impacts. In addition
to income poverty reduction, improved maize yields also have important implications for
household food security in those areas.

Finally, it is important to emphasize the need for complementary research in key
aspects that fall outside the scope of this essay but which are crucial for the advancement

of the sectors and for their sustained positive impact on broad based income growth and

 

69 Note that the notion of trade here includes both exports to the domestic (other parts of Mozambique) and
foreign (neighboring countries) rest of the world.

7° Episodes of trade restrictions, particularly in border areas, have been more common as local practice than
as an ofﬁcial central government policy; the latter is openly favorable to free trade. It is, therefore,
important that government be more active in ensuring that local authorities do not prohibit maize trade.

176

poverty reduction. We suggest that strategies that emphasize improved coordination for
facilitating investments in research and extension combined with area expansion and
increased productivity and quality at the farm level should be identiﬁed and
encouraged.71 Research needs to continue focusing on the analysis of the implications of
the current market and regulatory structure for competition and sector coordination, and
suggest ways to overcome current constraints to maximize the effects of interventions on

rural poverty.

 

7' That can be achieved with high yielding quality seeds, the strengthening of extension systems for
disseminating good ﬁeld practices and grading standards, and sufﬁciently remunerative producer prices to
ensure a continued and increasing participation of farmers in the sectors.

177

CHAPTER 5

SUMMARY OF FINDINGS AND POLICY IMPLICATIONS

5.1. Introduction

Poverty is widespread in Mozambique, with greater incidence in rural areas where
the highest proportion of the population live and work. Livelihood strategies among rural
households in the Zambezi Valley are predominantly based on agricultural activities, but
income diversiﬁcation is increasingly important. Cash income from agriculture comes
predominantly from tobacco and cotton production. Due to cash constraints and poor
access to inputs and credit by farmers, and high demand from buyers to meet quality and
volume requirements, contract faming is the dominant form in the organization of
transactions in those cash cropping sectors. As applied in Mozambique, government
grants processing companies legal monopsony rights to bliy all the output from
contracted farmers in speciﬁed geographical areas; in return, processors, who need to
spread their ﬁxed costs over as much product as possible, provide inputs on credit and
extension assistance to small farmers. Government additionally sets a minimum price for
seed cotton, though not tobacco, each year. The selective nature of contract farming
implies that not all households may have the chance to directly participate in these
schemes; some households are excluded. A key question, then, is how large and
widespread the indirect income effects of these schemes are, compared to the direct
effects. The answer to these questions has a lot to say about the poverty reduction effects
of such crops, and may generate insights about policies and programs to enhance these

effects.

178

The general purpose of this study is to help guide government and donors in their
decisions to allocate development resources to induce private sector investments that
yield high proﬁts to ﬁrms while generating broad based income growth and poverty
reduction. Speciﬁc research objectives are as follows. First, we seek to understand the
nature of contract farming schemes for cotton and tobacco, and the rationale for their
persistence. Second, we develop econometric models to assess the determinants of farmer
participation and proﬁtability in contract farming. Those models are extended to assess
the effects of participation on agricultural and total household incomes, accounting for
threshold effects of education and land holdings to identify more precisely the types of
farmers likely to beneﬁt more from contract farming. Finally, we use an economy-wide
model to account for feedback effects and evaluate the income poverty effects on
different household groups of expansion and exogenous shocks in cash cropping sectors.

This study uses data from a two-round survey undertaken with households in
tobacco and cotton concession areas of the Zambezi Valley of Mozambique. The study
also uses household consumption data from the National Expenditure survey (IAF 2002-
03) to derive household expenditure shares, and secondary data for the cotton and
tobacco sectors, both at the aggregate and ﬁrm level. The household level survey
collected data on the level of intermediate input use and the variation in factor use,
particularly seasonality in labor demand and household decisions with respect to the use
of family or wage labor and its allocation across competing activities. The survey also
collected data on production and marketing of crops, livestock, and ﬁshing, non-farm
enterprises and wage labor, asset ownership, and remittance income. The data collected

were used for the econometric analysis and for the construction of a Regional Social

179

Accounting Matrix for the Zambezi Vally (ZVR-SAM); this SAM served as the database
for a Regional Computable General Equilibrium (ZVR-CGE) model developed to
undertake policy simulations. In Section 5.2 we summarize the research ﬁndings. Section

5.3 presents a summary of policy implications and suggestions for future research.

5.2. Summary of Findings

In this Section we summarize the key research ﬁndings in this study. Therefore,
we start with the ﬁndings on the institutional analysis, followed by the econometric
analysis of determinants of participation and performance, and ﬁnally the economy-wide

model results on the impact of exogenous shocks on rural poverty.

5.2.1. Organization of Production and Trade in Cash'Cropping

The institutional/transaction costs analysis indicates that in both sectors contract
farming arrangements appear as an institutional response to widespread failure in input,
credit and output markets, and the lack of an effective public or private serviceprovision
network. The outgrower ﬁrms (ginners in cotton and multinational trading/exporting
ﬁrms with processing facilities in neighbouring countries in the tobacco sector) need to
ensure product quality and a large volume of purchases to reduce unit marketing and
processing costs. In general, farmers have very few cash generating alternatives, so that
input credit and a guaranteed output market appear as an important opportunity. Looking
at speciﬁc factors related to production, marketing and processing characteristics of the
crops, and a set of economic and political factors, the analysis concludes that some level
of non-market vertical coordination is likely to emerge under these circumstances as pure

spot markets are absent on both the input and the output sides and production

180

speciﬁcations require some degree of supervision and speciﬁc production techniques; this
is particularly important in tobacco, which uses a wider range of productivity enhancing
inputs and is more demanding in its ﬁeld practice requirements. The analysis ﬁnds that
full vertical integration, i.e., plantation type arrangements, is not feasible due to the labor
intensive nature of the production process that would make labor supervision costs
extremely high. Also, under a plantation type arrangement, the ﬁrms would be obliged to
pay the statutory agricultural minimum wage which is generally higher than the informal
wage that smallholders pay to hired labor and, in some cases, even higher than the
implicit wage that most smallholders end-up earning as contract cash crop growers .
While contract farming allows these systems to function, asymmetric information
and unbalanced bargaining power over issues such as prices and grading have emerged as
barriers to its development. Furthermore, especially in Cotton, low world market prices,
low ginning outtums, and low productivity at the farm level have constituted major
constraints. These are, in part, consequences of a wide range of market and coordination
failures and weaknesses in the concession system, many of which are beyond the scope
of this study; we focus ﬁrst on understanding the direct effects of the systems as they

currently exist, then examine economy-wide effects of expansion and exogenous shocks.

5.2.2. Econometric Analysis: Farmer Selection/Performance and Effects of
Participation
For each sub-sector we use a standard sample selection model to assess the

determinants of farmer participation and, to analyze the factors explaining the level of
proﬁts accruing to scheme participants. Then, we use a treatment effects model to assess
whether farmer participation in contract farming schemes has a signiﬁcant impact on

levels of crop and total household. To identify the types of farmers that beneﬁt from

181

participation, we investigate the effects of participation interacted with thresholds of
educational attainment and land holdings.72

Probit results for tobacco areas indicate that household participation in tobacco
contract farming schemes is more associated with endowments, technology, and income
diversiﬁcation opportunities and less with demographic characteristics. First, there are no
effects of household head gender, nor effects of educational attainment. Second, the
availability of draft animals, and the value of production tools increases the likelihood of
farmer participation. Third, there are no land threshold effects on participation. Finally,
households with greater livestock sales and wage labor are less likely to participate in
contract farming schemes.73

Tobacco growing households exhibit highly variable proﬁts; 30% lost money
during our survey year, while average annual proﬁts were $730.74, which represents
nearly half of the value of their total crop production. The analysis of the determinants of
tobacco cash income in the second step does not give evidence of sample selection bias.
Once households choose to engage in tobacco production, some effects are observed.
First, the level of education attained by the head of the household is not statistically
signiﬁcant. However, education plays an important role in determining access and
outcomes in off-farm income in tobacco growing areas. Regression analysis using a two-
stage procedure indicates that education of the household head is an important
determinant of participation in both self-employment and wage labor in tobacco growing

areas. However, only wage labor market eamings are statistically increased with the

educational attainment of the head (regression results are presented in Appendix B).

 

72 Education attainment and land area thresholds are also introduced in the proﬁts determinants model.
73 Each of the income diversiﬁcation variables was tested for endogeneity, using the Hausman Test. In each
case we failed to reject the null hypothesis of exogeneity.

182

Second, female headed households run less proﬁtable tobacco farming operations, with
proﬁts that are, on average, $400 lower than those. of their male counterparts. Third, land
holdings have a large effect on proﬁts only at the highest threshold level, with proﬁts Of
land rich tobacco growing households averaging $780 over that of land poor tobacco
growers; effects at lower levels of land holdings are not statistically signiﬁcant.74 Fourth,
the value of household production assets has a positive and statistically signiﬁcant effect
on earnings. Finally, agro-ecological/location speciﬁc ﬁxed-effects are observed; grower
households in mid-high altitude areas have proﬁts signiﬁcantly higher than those of their
counterparts in drier, lower altitude areas.

Model results for cotton areas in the ﬁrst stage probit indicate that the likelihood
of participation is not positively associated with the educational attainment of the
household head. This is consistent across cotton areas in Mozambique, where more
educated households show a higher propensity to engage in more proﬁtable non-farm
activities; two-stage regression analysis in cotton areas in the study region, however,
show that there is a positive association between education and participation in non-farm
self employment, but it is not statistically signiﬁcant. Likewise, a non-statistically
signiﬁcantrelationship is found between education and outcomes in non-farm activities.
Some other results stand out. First, households that have larger areas of land are more
likely to engage in a contract; all land threshold dummies are statistically signiﬁcant. This
is in sharp contrast with the tobacco results, which showed no impact of land holdings on

participation. A possible explanation for this difference is that, while land is the single

 

7‘ The ﬁnding of land threshold effects on performance but not participation suggests that access is not
biased towards the land rich. While access to the schemes is more driven by factors other than land
holdings alone (e.g., managerial ability), participation itself, all other factors equal, is more rewarding for
land rich smallholders.

183

most important factor for cotton production (under the current technological package),
participation in tobacco is more demanding, including the (unobserved) ability required
to manage production resources in a more complex set of ﬁeld activities. Second, unlike
in tobacco areas, a higher value of production and marketing assets has no effect on the
likelihood of participation in the schemes. Third, access to alternative sources of income,
such as livestock and self-employment, reduces the demand for cotton production
contracts. Due to the overwhelming use of family labor, wage labor income opportunities
do not compete with direct participation in contract farming.

Cotton growing households also exhibit variable proﬁts, with 20% incurring
losses; average proﬁts are only $93.60 per year, less than 20% of the value of their total
crop production. Second stage OLS regression results for farmer performance in cotton
schemes indicate the presence of sample selection bias. Among demographic variables,
only the number of adult equivalents is statistically signiﬁcant at 10%, each adding on
average $37 per year.75 Education of the household head is the only other demographic
variable close to signiﬁcant, which is a rough indication that, although highly educated
heads tend not to participate in cotton farming, when they do, they may do better than the
less educated ones. Furthermore, like in tobacco areas, total area owned is statistically
signiﬁcant only at the highest land holding quartile, where land rich cotton smallholders
exhibit proﬁts that are close to $150 higher than those of land poor cotton growers. Also,
the value of production and marketing equipment increases returns to cotton. Finally,
proﬁts in all Dunavant locations are statistically lower than those in Gorongosa , a C .N.A.

area.

\

 

7’ This result points to the relatively greater importance of family labor compared to hired labor in cotton
areas. In tobacco areas, given the importance of wage labor, the availability of adults in the household is
less important.

184

The Treatment Effects Model is designed to analyze the effect of participation
associated with education attainment and land holding thresholds, on the levels of crop
and total household income, controlling for other factors.

The model performs well in both areas and reveals some important results. In
tobacco areas there are indications of signiﬁcant returns to participation, but only at the
highest land holdings threshold for both crop and total household income levels”;
interaction effects of participation and land area owned dummies are only statistically
signiﬁcant and sizable ($1,306) at the highest land quartile. The magnitude and
signiﬁcance at that level is stronger for total household income ($1,576) suggesting that
even larger farmers appear to not be giving up on proﬁtable off-farm income generating
opportunities. There are no participation-education threshold effects. This is a surprising
result for a crop that requires careful management and which features steep price
discounts for poor quality; further investigation is warranted. The model reveals that
female headed households in tobacco areas earn lower crop incomes than their male
counterparts ($488 less). However, differences in total income are not statistically
different, which may suggest that off-farm income contributes to reduce gender
differentiation in incomes in those areas. Higher value of production and marketing assets
contributes to higher total household income. Finally, the model exhibits weak agro-
ecological or location ﬁxed effects. Likewise, incomes of participants across locations are
relatively balanced.

The treatment effects model results for cotton areas show no statistically

 

7‘ In tobacco growing areas, average total crop income at the household level is approximately $1,315
($1,573 among growers and $596 among non-growers), while total household incomes (crop plus income
from farm and non-farm activities off the household farm) is $1,606 ($1,815 among growers and $1,023
among non-growers).

185

signiﬁcant returns to participation, even when land holdings and education thresholds are
interacted with participation. This indicates that total crop and total household incomes77
between cotton growers and non-growers, after controlling for demographic, factor and
asset/technology endowments, and spatial factors, are not very different, and even income
across participants differentiated by land or education attainment are not very different.
Like in tobacco growing areas, but with an effect smaller in magnitude, higher value of

production and marketing assets contributes to increased crop and total household income

in cotton areas. There are no signiﬁcant district ﬁxed-effects in cotton areas.

5.2.3. Economy-wide Analysis: Income Poverty Effects of Expansion and Shocks

The econometric models gave some important indications regarding the
importance of landholdings in cash cropping and the identiﬁcation of the types of farmers
that beneﬁt from participation in the schemes.78 These results lead to the question of how
important the effects of expansion in cash cropping sectors are to different types of
households. The ZVR-CGE model introduced in the analysis was intended to further
investigate the effects on income growth and poverty reduction of cash crop expansion
and of a series of exogenous events hypothesised to take place during expansion.
Expansion was simulated as an increase in capital endowments speciﬁc to cash cropping
sectors. The other exogenous shocks simulated were selected on the basis of current
issues that are considered relevant for the sectors. They include increases in productivity,

increase in export prices for tobacco, cotton and maize, increase in import prices for

 

77 In cotton growing areas, averagehousehold total crop income is $479 ($518 for growers and $365 for
non-growers), while total household incomes (crop plus income from farm and non-farm activities off the
household farm) is $692 ($732 for growers and $574 for non-growers).

7’ Remember that land holdings threshold effects were statistically signiﬁcant for both cotton and tobacco
proﬁts. Land holding threshold effects on participation were only strong for tobacco growers.

186

inputs, and a government imposed tax on cash crop exports. The discussion of policy
implication in the next section contextualizes each of these shocks.

In the CGE model we assume full employment of activity-speciﬁc capital and
unemployment with full mobility of all other factors. Model simulations suggest that
growth rates of income and poverty reduction effects of expansion and exogenous shocks
are sizable in both areas, but larger in tobacco areas.

In tobacco areas, an expansion of 15% combined with a 15% increase in export
prices yields a greater impact than the same expansion with productivity gains of 15%. In
fact, expansion with higher world prices results in an increase of 49% in per capita
incomes of non-growers and 79% in the incomes of tobacco growers; expansion with
increased productivity results in relatively lower, but still signiﬁcant, growth rates of 43%
and 62%, respectively. Model results indicate that this growth in per capita incomes of
households (growers and non-growers) results from a number of linkage effects. Initial
expansion in the sector generates increased level of economic activity across all sectors,
including non-tobacco agricultural activities and non-farm activities. The relatively
higher level of economic activity results in greater remunerations for grower and non-
grower groups alike, via the increased demand of factors by those activities.

In contrast, in cotton growing areas, where productivity is extremely low,
expansion with productivity gains has a relatively greater impact than expansion with
increases in world market prices; growth in per capita incomes for cotton non-growers
and growers is 31% and 40%, and 21% and 24%, respectively. Model results indicate that
productivity gains (compared to price increases) in cotton areas generate a relatively

higher and broader (beyond cotton itself) growth in the level of economic activity across

187

all sectors. That translates in an increase in the volume of cotton that is proportionally
greater than the increase in price simulated, implying higher incomes to grower farmers.
The relatively higher level of economic activity results in‘greater remunerations for
grower and non-grower groups alike. It appears that the more efﬁcient use of labor in
cotton has important implications for other sectors as well. In fact, those activities are
able to use more labor as a result, i.e., they exhibit a greater growth (in percentage terms)
in labor demand although absolute base levels are typically higher in cotton production
itself. We conclude that productivity increases in the cotton sector lead to good prospects
for expansion in other sectors as well, and that it can potentially be beneﬁcial to non-
growers, even when beneﬁts to growers are limited.

Adverse events such as input import price increases and export taxes slow down
poverty reduction effects signiﬁcantly. The damages of increased input prices are more
severe in tobacco growing areas, where the input package is substantially more
expensive. The effects of an export tax are more severe in tobacco growing areas where it
signiﬁcantly slows down the effects of otherwise successful expansion efforts. In both
areas, better maize prices have very positive implications for poverty reduction.

The ZVR-CGE model results conﬁrm several indications from the econometric
analysis. First, effects of the cash crop on non-grower income growth and poverty
reduction, though lower relative to grower groups, are signiﬁcant. Greater demand for
factors in response to increased demand for farm and non-farm goods and services, and
its subsequent remuneration to household groups explains these broad based effects.
Second, relative effects in favour of grower groups are more accentuated in tobacco areas

than in cotton areas, conﬁrming that grower households in tobacco areas are not giving

188

.-h

up on proﬁtable non-farm opportunities that expand signiﬁcantly as cash crop production

expands.

5.3. Policy Implications and Future Research

In general, the cotton and tobacco sectors in Mozambique have provided a great
deal of the rural population in concession areas with a secure source of cash income in
areas where alternative income generating activities are limited. These cash crop sectors
are currently faced with a number of pressing issues. In this section we look at those
issues as they relate to the analysis presented in this study and the relevant implications
for poverty reduction in the study region. Such issues include the effects of a proposed
tobacco export tax on rural poverty; the implications of a cotton sector recovery that
relies on enhanced productivity; labor migration and growth of the tobacco sector;
interactions between rural education, cash cropping and poverty reduction; and a set of
general issues such as maize trade, input distribution and environmental and
technological spillovers.

First, a proposal to impose an export tax on raw tobacco, as a way of encouraging
domestic processing, has been the object of great controversy in recent years. Benﬁca et
al. (2005) discuss the implications of that policy in a cost-beneﬁt framework, and
conclude that the imposition of such a tax at this stage of the development of the tobacco
sector is not appropriate, and suggest that other ways to promote investments in
processing be found without compromising the long term sustainability of the sector.
Findings in this study strongly reinforce this recommendation: a 15% export tax would
eliminate any income gain to grower households from a 15% expansion of the tobacco

sector, and would reduce income gains to non-growers from 18% to 6%. It is, therefore,

189

recommended that Government not embark on such policy as it is not consistent with its
ultimate poverty reduction objectives.

Second, this study documents the low proﬁtability of cotton (associated
with low prices and low productivity) relative to tobacco in the Zambezi Valley
region. Part of the reason for this poor performance relates to the nature of the
cotton concession model as applied in Mozambique - which precludes
competition and does not balance this with any effective performance monitoring .
system. Therefore, we suggest that high priority is given to the development of a
more adequate management of the concession system in the near future.

In this study we ﬁnd that increased proﬁts in cotton can be achieved with
increased farm size and a higher level of production assets. However, there are no
landholding threshold effects on total crop income nor on total household income, which
suggest that the cotton activity generates some, but not yet very strong, economic
linkages that can sustain overall economic gains. This calls for policies aimed at higher
yields at the farm level and the promotion of non-farm businesses in those economies.
Economy-wide simulation results indicate that productivity gains in cotton have a broad-
based income growth and poverty reduction effect, even greater than increased world
prices; on a policy standpoint, because sustained increases in world prices are unlikely,
this result is encouraging. This study suggests that maximizing the contribution of cotton
to smallholder livelihoods will require increased productivity at the farm level, that can
come about through improvements in the seed stock (quantity and quality), as well as in
the input package, extension, and prices to farmers. Also, reforms currently underway

need to be participatory and be given top priority by stackeholders (public and private)

190

interested in the sector and in smallholder welfare, in order to better explore the available
alternatives.

Third, the successful expansion in tobacco production has been possible due to
the readily available labor in border areas that is knowledgeable of tobacco cropping.
Circumstantial evidence suggests that this labor force is a mixture of returned refugees
(established in Malawi during the Mozambique Civil War), family members of these
returnees, and a genuine new generation of Malawian migrant laborers, many of whom
are former tobacco smallholders in Malawi that ﬁnd the wage labor opportunities in
Mozambique more proﬁtable. In terms of our model, these patterns of employment raise
concerns about possible consumption leakages; the corollary in popular sentiment is that
this income should be going to Mozarnbicans, not Malawians. Yet, over 75% of the so
called “Malawians” report spending 9-12 months working in Mozambique, which
suggest that a great deal of their annual consumption takes place in Mozambique. In
practice our study suggests that (i) income leakage is not a major problem, and (ii)
availability of Malawian labor is important to the growth of the sector. A policy
implication is that efforts to ensure that Malawi migrants gain some kind of permanent
residency that leads them to spend more time and resources in the Mozambican territory
can be helpful to feed expansion of the sector and spread its beneﬁts in the local
economy.

Fourth, the lack of effect of education, while somewhat expected in cotton areas,
is surprising in a crop as demanding (in ﬁeld management) as tobacco. We suggest that
great scope remains for improving ﬁeld practices, yields, and proﬁtability; as companies

strengthen their extension efforts and more farmers have more time to learn and apply

191

improved techniques, more educated farmers will begin earning higher returns from
tobacco. The immediate policy implication is that incentives need to be put in place to
push companies to strengthen their research and extension systems and allow those
beneﬁts to cash crop farmers and consequently spread poverty reduction effects in rural
areas. More research in this area is needed.

A set of additional policy considerations are worth pointing out. First, maize is
important both as a food security crop and. a cash crop in these areas. Simulation results
indicated that better export prices for maize reduce poverty in all areas and across all
groups. Therefore, a continuation of a policy that does not restrict trade both in-country
and across national borders is important. Second, although results indicate limited
negative effects of high import prices for inputs, particularly in cotton areas, where the
input package is relatively cheaper, measures aimed at reducing the costs of importation
and transport are highly encouraged, as they can help minimize any negative effects from
factors outside the control of domestic agents. Finally, the issue of technological and
environmental spillovers requires more attention. On the positive side, survey results
indicate that the use of fertilizers on maize by non-growers of tobacco in tobacco areas is
very positive, given the strategic importance of that crop for poverty reduction and
household food security. On the negative side, tobacco expansion leads to extensive tree
cutting. Survey data indicate that the rate of tree cutting by tobacco growers far surpasses
the rate of planting. If those trends are not halted, the long term consequences can be
catastrophic. Speciﬁc actions to contain or reverse the situation are required.

Results in this research raise some issues that merit further investigation. One

such issue, particularly in tobacco growing areas, is related to the signiﬁcant effects of

192

-..u'

participant land holdings at the highest quartile and its implications for the relationship
between company expansion strategies (between one that focuses in a large number of
smallholders versus another that favors a smaller number of larger farmers) and poverty
reduction. On the one hand, we ﬁnd that, although proﬁtability is higher among those
with more land, overall, larger farmers are not more likely than others to cultivate
tobacco; in principle, one would expect farmers with more land wanting to get engaged in
the crop, and concession companies pushing towards more support to larger farmers.
Understanding why this is not happening is an important question to pursue through
further research. On the other hand, our ﬁndings indicate that tobacco growers are less
likely to engage in wage labor and many smaller tobacco farmers earn negative proﬁts
with the crop. In light of previous research indicating the occupational persistence of
wage earners, especially those on the high end of the labor market (Tschirley and
Benﬁca, 2002), and the increasing importance of wage labor in lifting rural incomes
(Boughton et al., 2005), an implication of the continued expansion of tobacco area is that
signiﬁcant wage labor employment opportunities will be generated, bringing higher
incomes to non-grower wage laborers. In this case, those left behind will be the smaller
cash crop growers, who, in addition to being more likely to have a loss in the crop, also
have little access to wage labor opportunities. Further research is necessary to better
understand the factors behind these results and get a clearer indication, for each sector, on
whether an expansion with lager farmers each getting relatively large levels of support is
favored over one that emphasizes small sized farms.

Finally, there is need for complementary research ‘in key aspects that fall outside

the scope of this study but which are crucial for the advancement of the two sectors and

193

4...-

for their sustained positive impact on broad based income growth and poverty reduction.

We suggest that strategies that emphasize improved coordination for facilitating
investments in research and extension, combined with area expansion and increased
productivity and quality at the farm level, should be encouraged. Furthermore, research
needs to continue focusing on the analysis of the implications of the current market

structure for competition and sector coordination, and suggest ways to overcome

constraints and maximize the effects of interventions.

194

."-’-

APPENDIX A
SAM MULTIPLIER ANALYSIS

The SAM (Social Accounting Matrix) based multiplier model evaluates the
effects of exogenous income injections (shocks) on the endogenous accounts as a result
of policy or program changes (Pyatt and Round, 1979 and 1985; Parikh and Thorbecke,

1996).

A balanced SAM requires that for each account incomes equal expenditures, i.e.,

2n,- =znk (3.1)
I j

where k refers to a given row I and column j corresponding to the same SAM account.

Formally, for a SAM multiplier analysis, the SAM transactions matrix, T, is

converted into a matrix of average shares (or average expenditure propensities), An,79 by
dividing the cells in each row (TU) by the column sums (yj):

Au = Tu / yj (8.2)

where z Ai,j = 1, for everyj.
i

Deleting the rows and columns for the exogenous accounts from An yields a sub-

matrix of endogenous shares, C“, with the entries Cn,J.8o The total income can now be

 

79 This shares matrix provides raw material for much economic analysis. The intermediate input
coefﬁcients in cell A2, correspond to the Leontief input-output coefﬁcients. The coefﬁcients for primary
factors in cell A3, are value added coefﬁcients and give the factor distribution'of income. Column
coefﬁcients for commodity accounts represent domestic and import shares, while those for households,
government and investment represent expenditure shares for these ﬁnal demanders. This matrix is the base
for the multiplier model explained in this section. This matrix is also provides the starting point for
estimating parameters of non-linear, neoclassical production functions, factor demand functions and
household expenditure systems of the type introduced in the CGE model in the next section.

80 Note that using average expenditure shares implies unitary expenditure elasticities, and hence the
computed average expenditure propensities are constant over any incremental exogenous injection, which
constitutes a major limitation. One mechanism used to relax this restriction is to incorporate marginal
rather than average expenditure propensities in the C matrix prior to computing the M matrix, with the help

195

computed as
Y" = C" Y" + X (B.3)
where X is exogenous income.

From this, it follows that

alYn = CndYn + dX (B.4)

dYn = M ch (B.5)

where the square matrix Mc = (I — C,,)'1 is the multiplier matrix, the inverse of the
‘identity matrix (1) less the SAM coefﬁcients matrix of the endogenous accounts’.

The Multiplier, MC, contains production effects in the rows corresponding to the
sectors and the income effects in the rows corresponding to factors and households.
Consider the multiplier matrix with entries MciJ, where Iprepresents the rows and j the
columns. Since Mc is a square matrix, the number of i’s and j’s is the same. The effects
of an injection or increase in exogenous demand for a sector — the effect of one additional
dollar - is given by the entry corresponding to that sector, i.e., in the main diagonal i=j.
The effect of that same dollar on the production of other sectors is found in that same

column j, but in the rows I that refer to those other sectors, and can be interpreted as the

production linkage effect. Total production multipliers can be estimated by summing the
multiplier of all sectors (MciJ) for the given j column of the impacted sector. The impact

on individual household groups can be read from the individual rows and the total effect
is the sum of the individual impacts — read vertically in the column corresponding to the

impacted sector. Likewise, the impact of transfers to the household groups can be

 

of consumer expenditure survey data (Thorbecke and Jung, 1996; Parikh and Thorbecke, 1996). Lack of
detailed information on household consumption behavior in mral areas may prevent that.

196

inferred by reading the multipliers, in the respective column of the impacted group, for
the different sectors of activity, factors and households.

More generally, therefore, the change in the incomes of the endogenous accounts

(dYn) — a column vector — can be derived by pre-multiplying the exogenous change (dX)

— a column vector — by the SAM multiplier matrix (MC) — a square matrix. Note that this

a multiplication of a matrix with a vector, and that the vector of changes has a number of
rows corresponding to the number of endogenous accounts and the entries will be non-
zero for the accounts that are being injected and zero for the others.

Also, there are leakages (L) in this process, i.e., part of the additional income

(dYn) will be leaked through induced demand for imports, induced government revenues

or induced savings. This can be represented as dL = Bde where B represents the
leakage propensity of the economy.
Table A.1 summarizes the SAM multiplier process. Yn is the vector of total

income or expenditures of the endogenous accounts; X is the sum vector of the

expenditures of the exogenous accounts; L is the column vector of the exogenous
accounts, Cn is the square matrix (n.n) of average shares of the endogenous accounts; and
B is the rectangular matrix (m.n) of the coefﬁcients with exogenous accounts as rows and

endogenous accounts as columns. The lower part of the table deﬁnes the important

concepts.8|

The regional multiplier matrix represents estimated total, direct and indirect

 

81 . . . .
All the endogenous accounts have to balance at a new equrlrbnum. Since the SAM as a whole as to
balance, the sum of the exogenous accounts also have to balance.

197

effects of exogenous income injections on the endogenous accounts of the regional SAM.
The regional input-output (Leontief) multiplier matrix (ML) is one component of the MC
matrix, which captures production linkages between economic activities that take place in
the region. In addition, the Mc matrix also captures expenditure linkages induced by
changes in production activities through their effect on incomes in the region. These
expenditure linkages are typically stronger than production linkages in many rural Sub-
Saharan African contexts (Haggblade et a1, 1987, Delgado et a1, 1998). To test the
relative strength of consumption linkages as opposed to production linkages, this study
will decompose the sectoral value added multipliers into direct, indirect and induced

effects.82

Table A.1. Endogenous and Exogenous Sub-Matrices

 

 

 

 

 

 

 

Endogenous Exogenous
1 Activities Endogenous
2 Commodities Transactions N Injections X Y
3 Factors (CnYn) X n
4 Households
5 Government Leakages Exogenous
6 Investment (BLYn) L Transactions t Y
7 Rest of Region x
Totals Y,n Y,x
29mm:
Mc = (I - Cn)’l Matrix of multipliers
dX Vector of shocks
dYn = Mch Vector of impacts
dL = BdYn The leakages

 

Source: Adapted from Pyatt and Round (1979).

These SAM based ﬁxed-price multiplier model assumes that all the endogenous

 

82 The so-called induced effect is the result of including household income and expenditure linkages in
multiplier estimation and can be taken as a measure of the strength of the consumption linkages. The direct
effect is simply the value added per unit of output and the indirect effect is the residual.

198

sectors have unlimited capacity to supply goods and services, i.e., an inﬁnitely elastic
supply - a Keynesian demand-driven system without resource constrains (Sadoulet and dc
J anvry, 1995, Taylor and Adelman, 1996). In order for the full multiplier effects to occur,
excess capacity and unused resources must prevail so that prices don’t change with
changing levels of output. Such an assumption may not hold true for agricultural
activities in many developing countries (Thorbecke, 1998). If constraints on regional
resources such as land, labor, seeds or chemical inputs (e.g., as a result of market
imperfections) are binding, they need to be taken into account when modeling the impact
of exogenous changes. Subramanian and Sadoulet (1990) and Lewis and Thorbecke
(1992) deal with the issue by imposing constraints on the production of selected sectors
in the form of perfectly inelastic supply. Parikh and Thorbecke (1996), in what they call
mixed multipliers, allow for the possibility of inelastic supply response beyond
predetermined output levels in selected sectors.83 The logic underlying those modiﬁed
multipliers, following Parikh and Thorbecke (1996), is as follows. Production capacity

levels are deﬁned for sectors assumed to be constrained. While excess capacity is

available in the constrained sector, the ﬁxed-price multiplier, MC, holds. Once the

deﬁned capacity is reached, the mixed multiplier, Mm, can be used for the remaining

demand.

 

83 Haggblade et al.(l99l) develop a price endogenous model to compare results with standard ﬁxed-price
models (Semi-Input-Output, Input-output and Economic Base). The model relaxes two key simplifying
assumptions of the standard multiplier models: (i) it accommodates an upward-sloping supply curve for
non-tradables, and (ii) allows for substitution among inputs (without imposing any speciﬁc functional form
on production) rather than insisting on a ﬁxed-coefﬁcient Leontief production technology. They conclude
that price endogeneity reduces multipliers substantially.

199

APPENDIX B

SELECTED OUTPUT TABLES

Table B.1. Determinants of MSE Income, Two Stage Model in Tobacco Areas

 

Explanatory Variables

Parameter Estimates

 

1St Stage: Has MSE Income? V

2"d Stage: Net MSE Income

 

 

 

 

 

Coeff 2 P > 12 l 1.32 Coeff t-stat P > | t 1
Contract Farming 0.251 0.88 0.38 -532.02 1.33 0.19
Demographics 3/
Female headed household -0.452 0.90 0.37 -334.92 1.59 0.12
Age of household head 0.036 2.47 0.01 ‘”" 0.27 0.05 0.96
Labor adult equivalents -0.215 1.76 0.08 -44.73 0.74 0.46
Education: 1-3 years 1.029 2.76 0.01 “”" 95.56 0.42 0.67
Education >3 years 0.909 2.52 0.01 " 538.93 1.51 0.13
Assets and Technology 4’
Area_QZ 1.009 2.72 0.01 * *
Area_Q3 0.319 0.73 0.45
Area_Q4 0.208 0.42 0.67
Use of Animal traction -0.763 1.48 0.14
Value of manual tools -0.003 0.64 0.52
Value of other equipment -0.001 0.42 0.68
Fertilizer in Maize 0.007 0.02 0.98
Agra-Ecological Eﬂects
Mid-Altitude
Angonia/MLT -o.321 0.64 0.52 1 1 1.93 0.72 0.47
Mualadzi/DIMON -0.134 0.28 0.78 733.94 1.37 0.18
Macanga/MLT 0.161 0.33 0.74 -99.95 0.40 0.69
Lower Altitude
Maravia/MLT - 0.477 0.82 0.42 -161.37 0.64 0.52
Luia/DIMON (dropped)
Inverse Mills Ratio ( 2) '130-79 031 0.76
Constant -1.312 1.65 0.10 411.82 0.70 0.48
Number of observations 159 94
Wald ch12 (17) 30.60
Prob > chi2 0.02
Pseudo R2 0.17
Log pseudo-likelihood - 90.24
F (11, 82) 0.60
Prob > F 0.82
R - Squared 0.25
Root MSE 842.08

 

” Probit equation for participation, 1 if has MSE income, 0 otherwise. 2’ Level of signiﬁcance (L8): 1 10%,
“' 5%, " 1%. 3’No schooling (Education=0) is excluded. 4’ Quartile l (Area_Ql) is excluded. MSE
income and value of assets are expressed in $US.

Source: Zambezi Valley Tobacco Concession Areas Study, 2004.

200

Table 8.2. Determinants of Wage Labor Income, Two Stage Model -Tobacco Areas

 

Explanatory Variables

Parameter Estimates

 

lSt Stage: Has Wage Labor? 17

2'“I Stage: Wage Labor Income

 

 

 

 

 

Coeff 2 P > | z[ LS2 Coeff t-stat P > | t 1 LS2
Contract Farming -0.781 2.55 0.01 ‘”“ 144.11 0.95 0.35
Demographics 3’
Female headed household 1.713 3.06 0.00 “ 437.26 2.82 0.01 "
Age of household head -0.018 1.16 0.25 2.68 0.58 0.56
Labor adult equivalents -0.276 2.14 0.03 "' 58.88 1.42 0.17
Education: 1-3 years 0.122 0.30 0.76 52.76 0.69 0.49
Education >3 years 0.715 1.79 0.07 * 408.92 2.71 0.01 ‘*
Assets and Technology 4/
Area_Q2 -0.715 1.79 0.07 *
Area_QB 0.116 0.27 0.79
Area_Q4 0.584 1.19 0.23
Use of Animal traction 0.324 0.52 0.60
Value of manual tools -0.004 0.81 0.42
Value of other equipment 000] 0.34 0.74
Fertilizer in Maize 0.435 1.16 0.25
A gro-Ecological Eﬂects
Mid-A Ititude
Angénia/MLT - 0.277 0.61 0.54 -200.97 1.41 0.16
Mualadzi/DIMON -l.454 2.71 0.01 " -125.97 0.77 0.44
Macanga/MLT -0.786 1.57 0.12 —313.96 2.04 0.05 *
Lower Altitude
Maravia/MLT -0.636 1.17 0.24 -419.57 2.59 0.01 "
Luia/DIMON (dropped)
Inverse Mills Ratio ( A. ) '30-36 0-19 0-35
Constant 1.830 2.19 0.03 "' -130.39 0.77 0.45
Number of observations 159 47
Wald chi2 (l 7) 49.12
Prob > chi2 0.00
Pseudo R2 0.29
Log pseudo-likelihood -76.27
F (11, 35) 5.90
Prob > F 0.00
R - Squared 0.53
Root MSE 254.24

 

l’Probit equation for participation, 1 if has Wage Labor income, 0 otherwise. 2’ Level of signiﬁcance (LS): l
10%, " 5%, " 1%. ’No schooling (Education=0) is excluded. 4’Quartile l (Area_Ql) is excluded. Wage
labor income and value of assets are expressed in $US.

Source: Zambezi Valley Tobacco Concession Areas Study, 2004.

201

r;

Table B.3. Determinants of MSE Income, Two Stgge Model in Cotton Areas

Parameter Estimates
2“d Stage: Net MSE Income

Explanatory Variables

lSt Stage: Has MSE? U

 

 

 

 

Coeff 2 P > 1 z 1 LSr Coeff t-stat P >11 1 L57
Contract Farming -1.057 3.56 0.00 *" 48.87 0.90 0.37
Demographics 3’
Female headed household -0.896 1.34 0.18 62.70 1.06 0.30
Age of household head -0.005 0.40 0.69 -0.09 0.06 0.95
Labor adult equivalents -0.155 1.43 0.15 -48.51 1.08 0.29
Education: 1—3 years -0.007 0.02 0.99 -74.18 1.34 0.18
Education >3 years -0.453 1.08 0.28 37.90 0.74 0.46
Assets and Technologv 4”
Area_Q2 0.153 0.40 0.69
Area_Q3 0.911 2.16 0.03 *
Area_Q4 0.134 0.27 0.79
Use of Animal traction 0.502 0.67 0.51
Value of manual tools 0.022 1.22 0.22
Value of other equipment -0.002 0.80 0.43
A gro-Ecological Eﬂects
Chiuta/Dunavant 0.383 0.81 0.42 53.06 0.57 0.57
Chifunde/Dunavant 0.058 0.12 0.90 9.15 0.13 0.90
Caia/C.N.A. 0.071 0.17 0.87 143.65 1.40 0.17
Moatize/C.N.A. 0.542 1.16 0.25 -l.12 0.02 0.99
Maringue/C.N.A. -0.415 0.87 0.38 103.42 0.71 0.48
Gorongosa (dropped)
Inverse Mills Ratio (2. ) '93-55 0-77 0-45
Constant 1.725 2.30 0.02 * 209.13 1.93 0.06 +
Number of observations 1 17 79
Wald ch12 (17) 35.43
Prob > chi2 0.01
Pseudo R2 0.19
Log pseudo-likelihood - 55.94
F (12, 66) 0.78
Prob > F 0.67
R — Squared 0.21
Root MSE 189.41

 

" Probit equation for participation, 1 if has MSE income, 0 otherwise. 2’ Level of signiﬁcance (LS): I 10%,
* 5%, ** 1%. 3’No schooling (Education=0) is excluded. 4’Quartile l (Area_Ql) is excluded. MSE
income and value of assets are expressed in $US.

Source: Zambezi Valley Tobacco Concession Areas Study, 2004.

202

Table B.4. Determinants of Wage Labor Income, Two Stage Model - Cotton Areas

 

Parameter Estimates

 

 

 

 

Explanatory Variables 1" Stage: Has wage Labor? V 2"d Stage: Wage Labor Income
Coeff 2 P > 1 z 1 Ls2 Coeff t-stat P > 1 t 1 LsT
Contract Farming 0.117 0.35 0.73 -96.76 1.72 0.10 **
Demographics 3’
Female headed household 0.120 0.19 0.85 420.58 2.54 0.02 *
Age of household head -0.017 1.06 0.29 -8.41 2.92 0.01 "”’
Labor adult equivalents 0.042 0.33 0.74 65.23 3.42 0.00 **
Education: 1-3 years 0.388 0.86 0.39 -99.75 0.74 0.47
Education >3 years -0.070 0.14 0.89 -158.69 1.42 0.17
Assets and Technology 4’
Area_QZ -0.272 0.61 0.54
Area_Q3 -0.487 0.98 0.33
Area_Q4 -0.024 0.04 0.97
Use of Animal traction 0.595 0.75 0.45
Value of manual tools 0010 0.40 0.69
Value of other equipment -0.003 1.09 0.28
Agro-Ecological Effects
Chiuta/Dunavant 2.196 3.30 0.00 ** 402.11 1.42 0.17
Chifunde/Dunavant 2.539 3.56 0.00 "”" 466.82 1.47 0.15
Caia/C.N.A. 1.871 3.04 0.00 "”“ 352.73 1.34 0.19
Moatize/C.N.A. 1.666 2.65 0.01 “ 316.03 1.37 0.18
Gorongosa/CNA (dropped)
Inverse Mills Ratio (1 ) 241-23 1-31 0-20
Constant -1.261 1.29 0.20 -159.27 0.37 0.72
Number of observations 98 41
Wald chi2 (18) 25.61
Prob > ch12 0.0598
Pseudo R2 0.26
Log pseudo-likelihood .49.1 1
R — Squared 0.48
Root MSE 160.86

 

l’Probit equation for participation, 1 if has Wage Labor income, 0 otherwise. 2’Level of signiﬁcance (LS): 1

10%, * 5%, ** 1%. ’No schooling (Education=0) is excluded. ”Quartile l (Area_Ql) is excluded. Wage

Labor income and value of assets are expressed in $US.

Source: Zambezi Valley Tobacco Concession Areas Study, 2004.

203

Table B.5. Regional MACROSAM Structure (%)

 

 

 

 

 

 

 

Expenditures
(Receipts) Activities Commodities Factors Households S-I GOV ROW Total
Activities 74.6 66.8
(50.8) (49.2) (100.0)
Commodities 24.0 6.5 22.5 100.0 29.7 102.4
(35.3) (6.5) (24.4) (6.4) (1.6) (25.9) (100.0)
Factors 76.0
(100.0) (100.0)
Households 94.2 70.3 -2.5
(97.2) (3.4) (-0.6) (100.0)
S-l 10.7 -20.6
(182.0) (-82.0) (100.0)
GOV 20.7
(100.0) (100.0)
ROW 18.8 5.8
(74.4) (25.6) (100.0)
Total 100.0 100.0 100.0 100.0 100.0 100.0 100.0

 

Note: The following payments were netted out: HH to ROW, HH to GOV and ROW to Factors.

204

 

...ch Scam 32< 5.3.8.80 8038. can .8on >o..a> .NBEaN ”850m ..x; .5 can ._ £6. + Am... 352.25% ..o .o>o..

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

N... NV 5. cm 20.29330 .0 59:52
8... v _. _- 25 on: m2 8...- ..Nm 9% assesses... Eoioaegom
on... mod- .2... m.mm. mm... «N. .- NS. 9.8 2:85 Loan.— own?
3 8... w . .m Nénm 23 cm... m . ._ how ”.3 SEE con. 0E8... .8:..=u.._m< .oZ
. 8... oz ..w . m .23.. we... 2.... SN. 3:: 858 c2. 05852288162
.3 cod 2.». ENS. Oman v.6 ow. N»; méom 088:. _a.=..=u_.w< .oz
2. Ed wed ..oood ”on... .N... R. ENS «Nae MES... 29.8.5: .02
- 0E8... 22.830:
3 co... .o.m- ..mm 0.: .md 86. m.mm adv EoEonEm .33 own?
+ me... E... NS 0.: .1. a... mom- ode new EoEAoEEoEow
:6 omd- oéo Nmo cod no... odo Goo c3833..
Eooaon. - 3%
am... cud o.~. ..n - - o.o ed 838%“; 1 SEES...
w..o mm. Dam «...N - - ed CA. 3.32 1.35.5...
a: c. .ao.Eo:U..o SD
3. cod .3... x. I. o. . m em... 8.. No On wEmD E83". 1 Cone... Emacs—Eon.
43%
S6 we... A... ...v cad w ..o- ...n 0.0 cocoa; .nE.=< .o 8:
.1 S... wmd 9.8 of“ 2.0 3... mg. v.2 me E2553. .0 o=.a> v8.2.3.
. 86 EN can 0. m. who .0... a. N. a. : me 200% .253). .0 o:_m> 3:83.
3 S... vw.~ «we v. v I 5... «Wm c v ...N 8.88.. 1 82 .80...
wom|m<Eﬂ
mm... mm... m.m 5m om... on... m.m NM 2.8.953. :3;N - .33
«2 3o- 2: we. 2... 8... 2... ...8 as. 1 .5: 22.88: Eco ow<
am... no. m4. 9». had 3. .- wan 0v Emu» 1 3.53. =1.... 5.333.
mm... 8.. N... ...N we... ww. .- 9N Wm Emu» 1 use... :1 05.? 5.30:3—
:5 3. .- ..n o. : on... M. _ .o- ...m he “:85“. 1 8.2.88: BB»: 295...
Que moo- ow 9o Rd 2.. ed v6 85 22.83:
a . . 4L .4 L
“3 _ . _ A n. 2%.. e965 29:80-82 ”3 _ . _ A A. Em; 2365 E58982
858....5 2....o ~32? 53E. 8:80....5 2.. .32.; 525
853.5% 32.335 @653“. .o 2.3. .c 858555 $26.56 20:5... .0 99:
m8}. 0830.. 89E .8on

 

mEermseﬂ ..oEEEaEm zo=a> .nanaN he macaw—8932.0 6831i 6d 033—.

205

.VOON Java—um map—<4 Commvan—OU OUUQDOF ave—w 60300 ~AU=M> mNoDEQN ”Dauzom

 

 

 

 

 

 

 

 

 

3:: 92: 3x: . 3x: 35. 98. 92: 33
e8... 8... S... s... S. a... 3. .553
S no 2 ..c c... .3 3. .<.z.u
2 2. 2 ..e o... 3. o... 2025
m... 3 a... 3 S. o... S. 52
ml... N1. wl. 3 c... S. o... 820 58 .8 eonxmhssxaz
3 am 3 E a: 3 a... .25
mm 3. 2 a... S. o... c... 55538
3 3.. 3. as 3” N. 3. was“:
3|. «ﬂ sin a 31. A o... m181E. om
...N a... Z 2 3 3 o... 350
2 2. ..o o... 2. ...N S. swag”...
m _ o. no 2 ...N 2 o... $8“.
mlw ad w.1.. min .31. m a... ”5.382%:
I 3: N... 2 2: 2. c... was:
3 S. 2 S n: 3 _ 3. .833...
31. a J aim g 31. c... 35.5 550
a: S: 3 3: 3 3: we 8.80-82
3. 2 _ N... no. 8 as 2. 8025-82
H 31m al. a 31. NAM a was“. 85 $6.52
2 3 2 Z a... 3 3 29,25
2: as «.2 2 a... 2: ...N. .<.z.u
31.. Md 31. 3 a a g 1111123“. 8:81
3 3 ca 3: 3 S 2: 2025
EN ”.2 a. _ m ...N». 2: SN 2:. 5:
H H ”.12. a Q H 3 was. 1... .o. 0.0321
«mom umom
was mots: 3.05.5-2; ES 535: 3022.72“.
.SEmU 11 .593 not: 8%.. N55”. 98..
.92 .83 822.2
.393“.

 

b333,. .3 233m $9.3. 3.3.—m ..Soah gum «Ssh.

206

6ch item 302 538280 8238. van .8on >o=m> ﬂop—5N ”850m

 

 

 

 

 

 

 

odo— odo— o2: ode— o.oo_ odo— odo— odo— £88m =<
Nd. o6 _.m_ ad 2: vs as _.n 3:93
ad ad we. Wu Z _ v.3 ad ”.1 “won—Ectmota:
mo 3 ad EN 3: 3 C.NN v.2 $3822-95
ad. CM. 5.: ”.mm Sum m.mm m._m Nmm nob:
adm mSm 95 a6 M: v.2 3 Wm Eon—Regain:
_.om odv ..om m4: NNM m.mm o. _ N 5.3 wczmotucbi
o. E 9.0» ﬁom ~.m~ odv Eon asm NNN 3:5“—
wdw 92: :m 93 ﬁve ode wdm Y? .883
ad ad 9mm _.o.r. 9mm mdm QNM m.mm 3:5
5on 8230b
wcimi V3883; .52 .52 2523.5 .<.Z.U ZOE—D HA2 £98“.
mszmm 25m“— 8080b
basin 850 no.0 :mmUéoZ map—.8; :otou

 

$223

mo£>29< act—m:— 1—8 «33834 ..a..5_=o_..w< "5333‘ no ..Soom £5?» 3.3.5 .833— .wd 039—.

207

.33. Scam 32< 5:68:00 8239—. new cocou Ao=~> ﬁanEnN ”venom

$3». ”396 3 as ”9%.? @2822 Ea .5 3.822 25 5 3%.: $533542“. 8:: Q
3025 >=E£ 02 A: :23 coEzmmm 22$ wEEEu 88:50 E vowmwco 333:8 .598 was mica; ”802

 

 

 

 

 

 

92: 3x: 92: 2:: :2: 3:: 328m =<

3 od 2m Q: ed 3 .8113 8
mi gm ”.8 ow 9o 3. 3935505:
3. a? ZN _.~ No we“ $532-5
2: 3:: 2m .. E 3 :2 . 32$
:3 od 3 9? 2m 2; 583585:
EN .3 3 3m 5. 3 ”.2 $5022-05
”.3 ed 3. 35 NS om ési
Ia 2:: So a? 3A. Ea .83
3 3 ed 3. 3 ed 23

850 225.850 mE—um; EEO muwﬁgom muoon.

moetom wctaoéscaz £29m“—
82>E<

 

moE>to< uni—m: 1:: «.9383»— ..a.:=_=u_..w< ub_>=u< ..o .835 £52» 8.3.5 532$ dd 03:.

208

.83 >02st van vach-2<m >N 2.: Bot 359.50 ”083m

 

 

 

 

 

odo— o.oo_ odo— o.oo_ odo— odo— odo— ode 0.2: 30,—.
adv ”no 06— new v.3“ 5mm 0.3“ xém N.mm 32580
”2 odm ..Nm Q: QNN Wom 542 Q: WE $2590.52

@8st SS 9.8. m...N..m -. -1-.. N8 mum -.vaizmSimmoE mEBEO 823$ -
Oo— w._ m. _ m w.m_ m. 2.. C.NN QB 06— m.o_ £0380
QMN m6 _.om Nuvm m. _ N x. _N ﬁom ﬁwm No— 30380-52
mdm _.N m. _ m m.N.m ﬁmv ”.mm m. ; m.mv m.mm mmo.< wEBEO 5500

“won .82.“: ”mom new “main:
BOM— >Omv 3:an ﬁcm “meta: Ivan umotm: uohnw Unaw—
38 322.2
$0.7.ch 392$

 

89C. 22.3.5: 398a .832:me 2:3:— LSoah .c—d oEah

209

.33 .325 98 383-33% >N 2: ES... c83EoU ”8.50m
.o\om.wm “3:93 A3 93 ”$9? ”33822 “mom can wczmotm: 2:.— AS "foo.”— nwczmotaséa 3:: ANV
con—3 Nazca”— 02 A: :23 uoEsmmm 22$ win—.8 82:80 E vowawco 8238a toaxo can maﬁa... ”Boz

 

 

 

 

 

 

 

3x: 92: 3:: 92: 92: 9555 .25 __<
9o- o. _- mo- 2. :3 - 882.com

3 3 nm 3 3 tea; 2055080
2: a: 2. 3 N2: .956
2: 3 1: ”.2 3 .moEwsaoE:
3; we 98 2: 92 €532-05
”3 m. a 3m 3m 3N RS:
3. S; v: of «.2 momNmsmotm:
3m 2m 3m as E 3532-2;
EM 3% 3m 3m v.2. ANENQ
NS 3m 33 0.8 :3 .33
EA Em 3. 3m 0: ES

mEozomzoI 3039.0 303050-:02 $530.0 muu>>ohmvu=oz
:< mEoﬁmsoz m2< 8030b $35825 m2< cocou
Boa Ea >00
15 088.: 29.8.8: 5 ﬂaws“; can 382$ .No mohmsm Eat whommcaﬁ. 82 was 392$

 

AX; 2:35 22328: E 2339:. can 282$ .8 mob—Em ._ —.m 93:.

210

 

 

 

 

 

doom Swim £85 :23850 823°... v5 :ocoU Ao=a> ion—5N ”350m

 

 

 

 

 

 

2033.530
EN 3 E a. E 5 2 go .352
5 _ N 2: 2: m SE SE 3.2 3:: 6:2 5.58 km
92: is: 92: ~23 3:: as: 3:: ”.22 c2: SS 98. :5 Ex: 2% 5:55: .58
2:8:— 20:83:
:5 33 2.0 :5 2.: 3.5 83 6.3 :3 8.2V 3.0 3.8V 3.9 2+8 2% .3
Z Z. Z 2: 3 2. 2 2: M: Ni 2 no. 3 n: 8280146
8.3 EE as 3.03 :6 3.3 5.: 33 as :2: 3.3 A3: 2.: 3.8 .ozuscszeom .o
...N Q3 2 OS 2 EN 3 35 2 3m 3 2: 9o 2. 222820.85 .m
S a: E So 3 wow 3: 22 S 9% 3 NS 3. 0.8 .82 ”was .v
3 as M: 2: 2 25 3 9.3: 3 22 N.“ in 2: in 255335928 .m
3 .3 3 a; 3 :5 3 03 v : as I: ...3 N: nﬂ £8.35 .N
am 33 SN 23 3m :3 Q: o.o a: f: 6.2 99: 3 3 act :20 .N._
Na 93 Na 5.. M: :m S I.» 3 3m x.“ ”...N «a a: .08. 28 .~._._
7%. SS 2% a?» «.3 S a as :2 o2 :3 a. _ m 23. 0% $3. Be 855324..
3m 3: 2;. 33 2:. 33 x 2 3: 3m 3: 3m 2: 20 2.3 32» noon .2
I: 3:; 2w 32.. v3 2:: E: w E n: 33 3: 38 as 2% 2:85 .2238? ._
.x. m3 ox: m3. .x. Q; ..\o ma .x. ma .x. Q; .x. E;
no.2 :< E9595 $0320-52 m2< __< $9520 $0395-52
:Ewum ‘52:; acuzanoU 0E8:—
iunEaN :< v.52 wEBEO 8238. 32¢. MEBEO c880

 

ﬁnes—3”.— ..oEc—Eaﬁm zo=a> ESEaN 2: 5 2:3:— Ecﬁmacm a: 252:5 2:. .~—.m a_aah

 

211

Table B.l3. Economic Structure in the Base: Export and Import Shares

 

 

 

 

 

 

 

Shares of Exmrts Shares of Imports
Exported In Total In Marketed Imported In Total In Domestic
Commodities Exports Output Commodities Imports Demand (%)
(%) (%) (%)
Packed cotton 9.6 100.0 Rice 2.9 74.9
Packed tobacco 73 .7 100.0 Beans l .2 3 1.7
Maize grain 3.6 50.0 Fruits 0.4 41.9
Groundnuts 3 .2 87.5 Coconuts 0.6 100.0
Root crops 0.7 41.9 Meat - cow 6.5 71.8
Vegetables 2.6 45.0 Cooking oil 3.4 100.0
Meat - goat 2.3 32.6 Sugar 2.2 100.0
Meat - birds 0.6 28.6 Salt 2.] 100.0
Alcoholic drinks 1.9 29.6 Tea/Coffee 0.2 100.0
Other commodities 1.8 24.0 Seeds 2.3 100.0
Pesticides 1.5 100.0
Fertilizers 30.7 100.0
Other agro-inputs 2.6 100.0
Fuel 18.8 100.0
Textile/footwear 9.3 80.6
Metal/blacksmithing 4.] 77.3
Soap and hygiene 9.5 100.0
Imported drinks/tobacco 1.7 100.0
Total 100.0 32.] Total 100.0 27.4
By Sector By Sector
Agricultural _ 93.4 50.0 Agricultural 5.1 4.2
Non-Agricultural 6.6 5.3 Non-Agricultural 94.9 39.0
Total 100.0 32.1 Total 100.0 27.4

 

Source: Zambezi Valley Regional SAM, 2003/4.

212

REFERENCES

Angrist, J. (2001 ). Estimation of Limited Dependent Variable Models with Dummy
Endogenous Regressors: Simple Strategies for Empirical Practice. Journal of Business
and Economic Statistics, 29, 1, pp. 2-15.

Amdt, C. (2002). HIV/AIDS, Human Capital, and Economic Prospects for Mozambique.
International Food Policy Research Institute, TMD Discussion paper # 88.

Amdt, C. (2005). The Doha Trade Round and Mozambique. World Bank Policy Research
Working Paper No. 3717 http://ssm.com/abstract=821406

Ballard, C., D. Fullerton, J. Shoven, and J. Whalley (1985). A General Equilibrium
Model for Tax Policy Evaluation. University of Chicago Press.

Bardhan, P. (1989). The New Institutional Economics and Development Theory: A Brief
Critical Assessment. World Development 17, 1389 - 1395.

Bardhan, P. and C. Udry (1999). Development Microeconomics. Oxford University
Press.

Barnow, B., G. Cain, and A. Goldberger (1981). Issues in the Analysis of Selectivity Bias.
In E. Stromsdorfer and G. F arkas, eds., Evaluation Studies Review Annual, Vol. 5,
Beverly Hills: Sage Publications.

Benﬁca, R., D. Tschirley and L. Sambo (2002).The Impact of Agro Industrial Investments
on Poverty Reduction in Rural Mozambique. DAP/MSU,‘Research Report No. 51E.

Benﬁca, R. (2003). Agricultural and Agro-Industrial Investment Strategies, Broad-based
Income Growth and Poverty Reduction in Rural Mozambique: A Regional Economy-wide
Approach. Ph.D. Dissertation Research Proposal. Department of Agricultural Economics,
Michigan State University.

Benﬁca, R., A. Miguel, J. Zandamela, N. Sousa, D. Boughton, D. Tschirley, and H.
Marrule (2004). How to Avoid Killing the Chicken That Lays the Golden Eggs: An

Analysis of the Potential Impacts of An Export Tax on Raw Tobacco in Mozambique.
DAP/MSU, Flash No. 42p.

Benﬁca, R., A. Miguel, J. Zandamela and N. Sousa (2005).T he Economics of Smallholder
Households in Tobacco and Cotton Growing Areas of the Zambezi Valley of
Mozambique. DAP/MSU, Research Report No. 59E.

Boughton, D., D. Mather, D. Tschirley, T. Walker, B. Cunguara, E. Payongayong (2005).

What is driving Mozambique ’s rural income growth? A comparative analysis of the I 996
and 2002 national agricultural sample surveys. Draft DAP/MSU, Research Report.

213

Boughton, D., D. Mather and D. Tschirley (2004). Income Growth in Rural Mozambique
1996-2002: A Challenging Comparison - Interim results based on TIA 96 and TIA 2002.
Powerpoint presentation prepared for a technical discussion at The World Bank,
November 17, 2004.

Carrilho, J ., R. Benﬁca, D. Tschirley and D. Boughton (2002). Qua] 0 Pope] da

A gricultura Comercial Familiar no Desenvolvimento Rural e Reduca'o da Pobreza em
Mocambique? Paper presented by Joao Carrilho at a Seminar of the Associaccio
Mocambicana de C iéncias e Tecnologias.

Cook, M. and Chaddad, F. (2000). Agro-industrialization, of the Global Food Economy:
Bridging Development Economics and Agribusiness Research. Agricultural Economics
23 (3)207 - 218.

DAD 4.2 : Distributive Analysis/Analyse Distributive — User’s Manual (2003). Software
developed by Jean-Yves Duclos and Araar Abdelkrin, and Programmed by Araar
Abdelkrin and Carl Fortin (Université Laval — MIMAP Project)

de Janvry, A. and E. Sadoulet (1987). Agricultural Price Policy in General Equilibrium
Models: Results and Comparisons. AJAE, 231-246.

de Janvry, A. and E. Sadoulet (1999). Agricultural Price Policy, Employment, and
Migration in a Diversified Rural Economy: A Village-Town CGE Analysis from Mexico.
American Journal of Agricultural Economics (81) 653-622.

Deaton, A. (1997). The Analysis of Household Surveys: A Microeconometric Approach to
Development Policy. Published by The John Hopkins University Press for World Bank.

Decaluwé, B., A. Patry, L. Savard and E. Thorbecke (1999).'Poverty Analysis Within a
General Equilibrium Framework. Working Paper 9909, CREF A 99-06.

Delgado, C., J. Hopkins and V. Kelly (1998). Agricultural Growth Linkages in Sub-
Saharan Aﬁica. IF PR1, Research Report Number 107.

Delgado, C. (1999). Sources of Growth in Smallholder Agriculture in sub-Saharan

Africa: The Role of Vertical Integration of Smallholders with Processors and Marketers
of High-value Added Items. Agrekon 38 (1999): 165-189.

Dervis, K., J. de Melo and S. Robinson (1982). General Equilibrium Models for
Development Policy. Part II: Computable General Equilibrium Models. Cambridge
University Press.

Donovan, C. (2004). Agricultural Growth and Poverty Reduction in Mozambique. Paper

prepared for the World Bank Video-Conference “Attacking Poverty in Mozambique”
held in Washington DC and Maputo on January 13, 2004.

214

Dorward, A., J. Kydd, and C. Poulton (1998). Smallholder Cash Crop Production Under
Market Liberalization: A New Institutional Economics Perspective. CAB International.

Dorward, A. (2001). The Effects of Transaction Costs, Power and Risk on Contractual
Arrangements: A Conceptual Framework for Quantitative Analysis. Journal of
Agricultural Economics 25 (2), 59 - 73.

Gangopadhyay, S. and K. Sengupta (1987). Small Farmers, Money Lenders, and Trading
Activity. Oxford Economic Papers, 39.

Greene, W. (2003). Econometric Analysis. Fifth Edition. Prentice Hall.

Glover, D. (1984). Contract Farming and Smallholder Schemes in Less Developed
Countries. World Development Vol. 12 (1 1/ 12), pp. 1143-1157.

Glover, D. and K. Kusterer (1990). Small Farmers, Big Business: Contract Farming and
Rural Development. New York: St. Martin’s Press.

Goldsmith, A. (1985). The Private Sector and Rural Development: Can Agribusiness
Help the Small Farmer? World Development Vol. 13 (10/11), pp. 1125-1138.

Government of Mozambique (2001). Plano de Accc‘io para a Reducc‘io da Pobreza
Absoluta, 2001 -2005 (PARPA 1).

Government of Mozambique (2005). Plano de Accc’io para a Reducc'io da Pobreza
Absoluta, 2006—2009 (PARPA II).

Gugerty, M., and GP. Timmer (1999). Growth, Inequality and Poverty Alleviation:
Implications for Development Assistance. Mimeo.

Haggblade, S., P. Hazel] and J. Brown (1987). Farm/Non-farm Linkages in Rural Sub-
Saharan Africa: Empirical evidence and Policy Implications. Agriculture and Rural
Development Department.The World Bank.

Haggblade, S., J. Hammer and P. Hazel] (1991). Modeling Agricultural Growth
Multipliers. American Journal of Agricultural Economics, Vol. 73: 361 -74.

Hazel], P. (1984). Rural Growth Linkages and Rural Development Strategies. Presented

in the 4th European Congress of Agricultural Economics.

Heckman, J. (1979). Sample Selection Bias as a Speciﬁcation Error. Econometrica, Vol.
47, No.1 (Jan 1979), pp. (153-162).

Horus (2004). Estudo C omplementar para o Melhoramento da Estratégia no Sector do
Algodc'io em Mocambique. Draft.

215

Huppie, M. and M. Ravallion (1991). The Structural Structure of Poverty During an
Adjustment Period: Evidence for Indonesia in the Mid-803. World Development 19 (12).

Jaffee, S. and J. Morton, eds. (1995). Marketing Africa ’s High-value Foods:
Comparative Experiences of an Emergent Private Sector (Dubuque, Iowa, USA:
Kendall/Hunt Publishing);

IN E (2000). Mozambique 1997 Population Census.

Intriligator, M., R. Bodkin and C. Hsiao (1996). Econometric Models, Techniques, and
Applications, Second Edition. Prentice Hall. New Jersey.

Jaffee, S. and Morton, J ., eds. (1995). “Marketing Africa’s High-value Foods:
Comparative Experiences of an Emergent Private Sector”. (Dubuque, Iowa, USA:
Kendall/Hunt Publishing);

Key, N. and D. Runsten (1999). Contract Farming, Smallholders, and Rural
Development in Latin America: The Organization of Agro-processing Firms and the
Scale of Outgrower Production. World Development Vol. 27, No. 2, pp. 381-401.

Klein, P. (1995). Empirical Research in Transaction Costs Economics: A Review and
Assessment. The Journal of Law, Economics and Organization, Volume 11, number 2.

Lemaitre, P., Fok, M. and Jeje, J .J . (2001). Cotton Sub-sector Study. Final Report.
Maputo: IAM.

Lewis, B. and E. Thorbecke (1992). District Level Economic Linkages in Kenya:
Evidence Based on a Small Regional Social Accounting Matrix. World Development,

Vol. 20, No.6 (881-897).

9

Little, P. and M. Watts (1994). Living Under Contract: Contract Farming and Agrarian
Transformation in Sub-Saharan Africa. Madison: University of Wisconsin Press.

Lofgren, H., R. Harris and S. Robinson (2002). A Standard Computable General
Equilibrium (CGE) Model in GAMS. IFPRI - Microcomputers in Policy Research, No.5.

Lofgren, H., S. Robinson, and M. El-Said (2003). Poverty and Inequality Analysis in a
General Equilibrium Framework: The Representative Household Approach. Chapter 15
- The World Bank toolkit.

Lofgren, H., and S. Robinson (1999). Non-separable Household Decisions in a
computable General Equilibrium Model. American Journal of Agricultural Economics,
Vol. 81 (3): 663-670.

Maddala, G. (1983). Limited Dependent and Qualitative Variables in Econometrics. New
York: Cambridge University Press.

216

 

Minot, N. (1986). Contract Farming and its Effect on Small Farmers in Less Developed
Countries. Working Paper 31 (East Lansing, Michigan, USA: Michigan State
University).

MPF/IFPRI/Purdue University (2004). Poverty And Well-Being In Mozambique:
The Second National Assessment. Maputo-Mozambique.

North, D. (1990). Institutions, Institutional Change and Economic Performance.
Cambridge: Cambridge University Press.

Oﬁco, A. O. and Tschirley, D. (2003) An Overview of the Cotton Sector in Mozambique.
Mimeo.

Parikh, A. and E. Thorbecke (1996). Impact of Rural Industrialization on Village Life
and Economy: A Social Accounting Matrix Approach. Economic Development and
Cultural Change 44 (2): 351-377.

Poulton, C., P. Gibbon, B. Hanyani-Mlambo, J. Kydd, W. Maro, M. Larsen, A. Osorio,
D. Tschirley and B. Zulu (2004). Competition and Coordination in Liberalized African
Cotton Market Systems. World Development Vol. 32, No. 3, pp. 519—536.

Pyatt, G. and J. Round (1979). Accounting and Fixed Price Multipliers in a Social
Accounting Framework. Economic Journal Volume 89 (356): 850-873.

Pyatt, G. and J. Round (1985). Social Accounting Matrices: A Basis for Planning. The
World Bank, Washington.

Ravallion, M., and S. Chen (1997). What Can New Survey Data Tell Us About Recent
Changes in Distribution and Poverty. The World Bank Economic Review, 11.

Robinson, 8., A. Cattaneo and M. El-Said (1998). Estimating a Social Accounting Matrix
Using Cross Entropy Methods. Intemationa] Food Policy Research Institute, TMD
Discussion paper # 33.

Rodgers, G. (1986). The Theory of Output-Income Multipliers with Consumption
Linkages: An Application to Mauritania. PhD Dissertation, University of Wisconsin,
Madison, Wisconsin.

Sadoulet, E. and A. de Janvry (1995). Quantitative Development Policy Analysis. The
Johns Hopkins University Press.

Staatz, J. (1988). The New Agricultural Economics: Applications of Transaction Cost /
Institutional Analysis to Agricultural Economics. Seminar Presented to the Scientiﬁc
Society for Finnish Agriculture and the Faculty of Agriculture and Forestry, University of
Helsinki.

217

SataCorp. (2003). Stata Statistical Software: Release 8. 0. College Station, TX: Stata
Corporation.

Stone, R. (1978). The Disaggregation of the Household Sector In the National Accounts.
World Bank Conference on Social Accounting Methods In Development Planning,
Cambridge.

Strasberg, P. (1997).Smallholder Cash-Cropping, Food-Cropping and Food Security in
Northern Mozambique. Ph.D., Dissertation, Michigan State University, East Lansing,
Michigan.

Subramanian, S. and E. Sadoulet (1990). The Transmission of Production Fluctuations
and Technical Change in a Village Economy: A Social Accounting Matrix Approach.
Economic Development and Cultural Change 39 (1): 131-73.

Tarp, F., C. Arndt, H. Jensen, S. Robinson and R. Heltberg (2002). Facing the
Development Challenge in Mozambique: An Economywide Perspective. IFPRI Research
Report # 126.

Taylor, J. Edward, G. Dyer and A. Yunez-Naude (2005). Disaggregated Rural
Economywide Models for Policy Analysis. World Development, Volume xx, No. xx, xxx-
xxx (In the Press).

Taylor, J. E. and I. Adelman (1996). Village Economies: The Design, Estimation and Use
of Villagewide Economy Models. Cambridge University Press.

Taylor, J. E., A. Yunez-Naude, and G. Dyer (1999). Agricultural Price Policy,
Employment, and Migration in a Diversified Rural Economy: A Village-T own C GE
Analysis from Mexico. American Journal of Agricultural Economics, 81, 653-662.

Thorbecke, E. and HS, Jung (1996). A Multiplier Decomposition Method to Analyse
Poverty Alleviation. Journal of Development Economics 48: 279-300.

Thorbecke, E. (1998). Social Accounting Matrices and Social Accounting Analysis
(Chapter 7, in Methods of Interregional and Regional Analysis, by Isard, W., I. Azis, M.
Drennan, R. Miller, S. Salzrnan, and E. Thorbecke. Ashgate Publishing Company,
Brookﬁeld, VT.

Tschirley, D. and M. Weber (1994). Food security strategies under extremely adverse

conditions: The determinants of household income and consumption in rural
Mozambique. World Development, Vol. 22:2 (159-173).

Tschirley, D. and R. Benﬁca (2001). Smallholder Agriculture, Wage Labor and Rural

Poverty Alleviation in Land-A bundant Areas of Africa: Evidence From Mozambique. The
Journal of Modern African Studies, Vol. 39:2 (333-358).

218

 

{

Tschirley, D. (2002). Some Characteristics of Pro-Poor Growth and Policy Implications
for Mozambique. Flash No. 29E, Minisz of Agriculture and Rural Development,
Department of Policy Analysis, Mozambique.
http://www.aec.msu.edu/agecon/fSZ/mozambigule/ﬂash29e.pdf

Tschirley, D., A. Oﬁco, and D. Boughton (2005). Mozambique Country Study.
Collaborative Research Project on Competition and Coordination in Cotton Market
Systems in Southern and Eastern Africa.

Tschirley, David, Colin Poulton, and Duncan Boughton (2006). The Many Paths of
Cotton Sector Reform in Eastern and Southern Africa: Lessons from a Decade of
Experience. Forthcoming in Cotton, Natural Resources and Society in Sub-Saharan
Africa, William G. Moseley and Leslie C. Gray, Eds.

Walker, T., D. Tschirley, J. Low, M.Tanque, D. Boughtoir, E. Payongayong, and M.
Weber (2004). Determinants of Rural Income, Poverty and Perceived Well-Being in
Mozambique in 2001-2002. MADER/DE/DAP Research Report # 57E.

Warning, M. and W. H00 (2000). The Impact of Contract Farming on Income
Distribution: Theory and Evidence. Presented at the Western Economics Association
Intemationa] Annual Meetings.

Warning, M. and N. Key (2002). The Social Performance and the Distributional
Consequences of Contract Farming: An Equilibrium Analysis of the Arachide de Bouche
Program in Senegal. World Development Vol. 30, No. 2, pp. 255-263.

Williamson, 0. (1991). Comparative Economic Organization: The Analysis of Discrete
Structural Alternatives. Administrative Science Quarterly 36, 269 - 296.

Willis, R., and S. Rosen (1979). Education and Self- Selection. Journal of Political
Economy, 87, pp. S7 - S36.

Wooldridge, J. (2000). Introductory Econometrics: A Modern Approach. South Western
College Publishing.

Wooldridge, J. (1999). Econometric Analysis of Cross-Section and Panel Data.
Massachusetts Institute of Technology.

World Bank (2005). Study on Contract Farming and Supply Chain Financing in
Mozambique.

219

 

 

   

utgtttttttgttg.1151119131