r 7..
“aim... an?
Emu. ma... . _ .
.: ‘fi . . ‘
. 5.2.5. 1 . ‘ . . . ‘ . . $5.1. ‘ r
E A .V' V ‘ . ‘ n .
at. xum...uvlm.:
r ......ޣ..13.; a... w...
L“ .Wflh‘u‘vwdiwzwx.“

‘ . .a
. gr."

‘1‘: .n ‘ y ‘ 1?! h .‘ 7C .3!
x .y. ;:....u n .2.
. 1.-.?! .
, a:

.
.
Jiv :3”???
IL)... . ,.. y...
ftiVnuuuuf :l‘rkq. 1H ..
«mfiygaiut
456.4.
r 38.5.]...

 

2 .
1.37.9
1

I;
31.1....
.13: .L

 

4., . ‘ zfiutidfig .,
| ||

 

This is to certify that the
thesis entitled

ASSESSING THE POTENTIAL ECONOMIC BENEFITS OF
TRANSGENIC COTTON IN MOZAMBIQUE

presented by

Raul Oscar Pitoro

has been accepted towards fulfillment
of the requirements for the

MS. degree in Agricultural Economics

 

 

 

‘ ,
C/‘QQJ VMLA/[élr’x

<1 ----- ~~M or Profe‘s‘sor’s Signatt7e
(54%,?

Date

 

MSU is an Affirmative Action/Equal Opportunity Institution

 

LIBRARY
Michigan State
University

 

 

 

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

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

6/01 cJCIRC/DateDuepGS-sz

ASSESSING THE ECONOMIC BENEFITS OF TRANSGENIC COTTON
IN MOZAMBIQUE: EX-ANT E ANALYSIS

By

Raul Oscar Pitoro

A THESIS

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

MASTERS OF SCIENCE

Department of Agricultural Economics

2004

ABSTRACT

ASSESSING THE POTENTIAL ECONOMIC BENEFITS OF TRANSGENIC
COTTON IN MOZAMBIQUE

By
Raul Oscar Pitoro

This study analyzes the potential economic impacts from adopting Bt cotton in
Mozambique more specifically, (I) the benefits to farmers, (2) the benefits to the society.
In history of Mozambican agriculture, this study is the first attempt to evaluate the
potential economic benefits from adopting Bt cotton, expected to provide insights for
decision making related to Bt cotton in Mozambique and in other countries with
smallholder production of cotton that are likely to consider the adoption of Bt cotton in
Sub-Saharan Africa.

The study concludes that under current practices, specially regarding to lower
insecticide applications, Bt cotton is not expected to generate additional income to farmer
but it is expected to do so for the society. However, it can be attractive to farmers if its
yield increases by 6%, about 40 Kg/ha over the actual conventional cotton yield level.
Alternatively, the Bt seed price would have to decrease by 17% to $13.78/ha. Moreover,
if farmers were to pay full cost of pesticides, and get paid export parity price for cotton,
they could pay full cost of Bt seed and earn more profit with Bt cotton than with
conventional cotton. Finally, more focused research is still needed to fully assess the

economic, environmental, and social benefits and risks of Bt cotton.

DEDICATION

In memory of my mother, Esperanca Raul. You always believed that life has good things
that people can benefit from them as long as enough energy and self confidence are put to
overcome difficulties along the way. Recognizing that your lesson was very important to
achieve my goals, this is an ideal moment to express my gratitude. I'll always remember

you.

-m-

ACKNOWLEDGMENTS

I would like to address my special gratitude to people and organizations that contribute
direct on indirectly to this achievement. Prof. David Tschirley, my major professor and
assistantship supervisor, Prof. Scott Swinton, Prof. Karim Maredia, and Dr. Thom Walker
were very helpful in guiding me on this research from the research outline up to final

report.

Special appreciation goes also to Prof. Eric Crawford, Dr. Douglas Landis, and Dr.
Duncan Boughton at Michigan State University, Prof. John Sanders at Purdue University,
Eugene Eulitz, Dr. Luisa Santos and Dr. Marcos Freire at University Eduardo Mondlane,
Graham P. Head, and Wally Head from Monsanto for providing assistance on crucial

steps of the research.

I am also thankful to the Food Security H Project and the Institute of International
Education (HE/Fulbright) who sponsored my studies with financial assistance. In
Mozambique I did appreciate the technical collaboration of Maria J osé Texeira, Higino de

Marrule, Jaquelino Massingue, and Anabela Mabota.

My colleagues and lovely fiiends from Michigan State University are also part of my

gratitude list, including Rui Benfica, Marcia, Ruy Allan, Feliciano Mazuze, Danilo

Abdula, Elliot Mghenyi, Deniz Nazlov, Gerald Nyambane, David Mather, Kofi Nouve,

-iv-

Alyse, Martinez, Lilian Kirimi, Ragassa, Raymond, and Rafael Uaiene. Your comments

helped to keep focused on the central issues of my thesis.

To my brothers and cousins for your help when I most needed Derito, Dina, Zito,
Peancha, Vadinho, Emidio, Zeca, Raul. To Nilza Bibi Nhancale for your courage to
accept the challenge of staying always awake during the very bad times of this thesis, I

love you so much.

At the US. consulate in Mozambique my greetings goes to Argentina Magaia, Robin
Smith, and Sonia for all arrangements concerning my Visa and many more, many thanks

for you all.

TABLE OF CONTENTS

LIST OF TABLES ................................................................................... ix
LIST OF FIGURES .................................................................................. xi
KEY OF SIMBOLS AND ABBREVIATIONS ................................................. xii
CHAPTER ONE ................................................................................................................. 1
1.1 CONTEXT .......................................................................................................... l
1.2 OBJECTIVES ..................................................................................................... 3
1.3 JUSTIFICATION OF THE STUDY ................................................................... 4
1.4 ORGANIZATION OF THE STUDY .................................................................. 5
1.4 ORGANIZATION OF THE STUDY .................................................................. 5
CHAPTER TWO ................................................................................................................ 6
2.1 COTTON SUB-SECTOR IN MOZAMBIQUE .................................................. 6
2.1.1 STRUCTURE ................................................................................................. 6
2.1.2 PERFORMANCE ......................................................................................... 11
2.1.3 BIOSAFETY m MOZAMBIQUE ............................................................... 12
2.2. BT COTTON ..................................................................................................... 13
2.2.1 IMPACT OF Bt COTTON IN DEVELOPING AND DEVELOPED
COUNTRIES ............................................................................................................ 1 5
CHAPTER THREE .......................................................................................................... 29
3. l APPROACHES ................................................................................................ 29
3.2 PARTIAL BUDGET ......................................................................................... 32
3.3 YIELD ESTIMATION AND ASSUMPTIONS ................................................ 34
3.4 DATA SOURCES ............................................................................................. 36
3 .5 THEORETICAL EMPIRICAL MODEL .......................................................... 39
3.6 COTTON YIELD MODEL ............................................................................... 43
3.7 CHOICE OF FUNCTIONAL FORM FOR COTTON YIELD MODEL ........... 47
3 .8 FUNCTIONAL FORMS ................................................................................... 48

-Vi-

CHAPTER FOUR ............................................................................................................. 54

4.1 MODEL PROPERTIES .................................................................................... 54
4.2 EVALUATION AND SELECTION OF A YIELD MODEL ............................ 55
4.4 COTTON YIELD ESTIMATES ....................................................................... 66
CHAPTER FIVE ............................................................................................................... 72
5.1 PARTIAL BUDGET ......................................................................................... 72
5.1.1 RELEVANT ADDITIONAL COSTS .......................................................... 73
5.1.2 RELEVANT ADDITIONAL BENEFITS .................................................... 76
5.2 PARTIAL BUDGETS AND ANALYSIS ......................................................... 77
5.2.1 ANALYSIS OF PARTIAL Bt COTTON BUDGETS .................................. 77
5.2.2 SENSITIVITY ANALYSIS FOR FINANCIAL PROFITABILITY ........... 82
CHAPTER SIX ................................................................................................................. 85
6.1 ECONOMIC PARTIAL BUDGET ................................................................... 85
6.1.1 RELEVANT ADDITIONAL COSTS .......................................................... 87
6.1.2 RELEVANT ADDITIONAL BENEFITS .................................................... 87
6.1.3 NET ECONOMIC PROFIT FOR Bt COTTON ........................................... 90
6.2 SENSITIVITY ANALYSIS FOR SOCIAL PROFITABILITY ....................... 97
6.3 NET TRANSFERS ........................................................................................... 99
CHAPTER SEVEN ............................................................................... L ........................ 102
7.1 SUMMARY OF OBJECTIVES, FRAMEWORK, AND STUDY
CONTRIBUTION ....................................................................................................... 102
7.2 SUMMARY OF FINDINGS .......................................................................... 103
7.3 POLICY IMPLICATIONS .............................................................................. 105
7.4 LIMITATIONS OF THE STUDY ................................................................... 1 10
7.5 RECOMMENDATIONS FOR FURTHER RESEARCH ............................... l 1 1
APPENDIX A ............................................................................................................... 1 134
APPENDIX B ................................................................................................................. 1 17
APPENDIX C ................................................................................................................. 123
APPENDIX D ................................................................................................................. 125
APPENDIX E ................................................................................................................. 128

-vii-

APPENDIX F .................................................................................................................. 1 3 1

APPENDIX G ................................................................................................................. 133
APPENDIX H ................................................................................................................. 138
APPENDIX I ................................................................................................................... 141
APPENDIX J .................................................................................................................. 148
BIBLIOGRAPHY ........................................................................................................... 159

-viii-

TABLE 2-1

TABLE 3-1

TABLE 4-1

TABLE 4-2

TABLE 4-3

TABLE 4-4

TABLE 4-5

TABLE 5-1

TABLE 5-2

TABLE 6-1

TABLE 6—2

LIST OF TABLES

YIELD AND COST ADVANTAGES OF BT OVER NON-BT COTTON

IN SELECTED COUNTRIES. ................................................................. 16
SUMMARY STATISTICS OF THE VARIABLES USED IN THE
REGRESSION MODEL, IN NORTHERN MOZAMBIQUE .................. 49
ESTIMATED PRODUCTION FUNCTIONS IN NORTHERN
MOZAMBIQUE, 2003. ............................................................................ 56
CHOICE CRITERIA FOR COTTON YIELD MODEL ........................... 60
MEAN OBSERVED YIELD AND PREDICTED YIELD FROM THE

THREE MODELS IN KG/HA IN NORTHERN MOZAMBIQUE. ........ 61

MARGINAL PRODUCTS, MARGINAL VALUE PRODUCT, AND
MARGINAL FACTOR COST FOR BT COTTON PRODUCTION AT
MEAN VALUES OF INDEPENDENT VARIABLES. ........................... 66

ESTIMATED BT COTTON YIELD AND BT YIELD GAIN OVER
NON-BT COTTON BY SEEDING WEEK AND NUMBER OF
ADDITIONAL INSECTICIDE APPLICATIONS, IN KG/HA IN
NORTHERN MOZAMBIQUE, 2003. ..................................................... 67

BT COTTON FINANCIAL PARTIAL BUDGET BY SEEDING WEEK,
AND IN SECTICIDE APPLICATION, IN NORTHERN MOZAMBIQUE,
2003 ........................................................................................................... 80

ESTIMATED S WI I CHING VALUES FOR A REPRESENTATIVE
FARMER TO ACHIEVE 100% RETURNS TO INVESTMENT ON BT
COTTON, NORTHERN MOZAMBIQUE, 2003 .................................... 83

FARM-GATE ECONOMIC COST SUMMARY IN NORTHERN
MOZAMBIQUE, 2003. ............................................................................ 89

ECONOMIC RETURNS TO CONVENTIONAL AND BT COTTON IN
NORTHERN MOZAMBIQUE BY SEEDING WEEK AND
INSECTICIDE USE, 2003 ........................................................................ 92

-ix-

TABLE 6-3

TABLE 6-4

TABLE C-l

TABLE D-l

TABLE D-2

TABLE E-l

TABLE F-l

TABLE 11-]

TABLE I-l

TABLE [-2

TABLE [-3

TABLE [-4

TABLE [—5

TABLE J-l

TABLE J-2

SWITCHING VALUES BY INPUT USE LEVEL, UNDER TYPICAL
PRACTICE IN NORTHERN MOZAMBIQUE, 2003. ............................ 97

TOTAL NET TRANSFERS IN THE NATIONAL ECONOMY WI I H BT
COTTON PRODUCTION, 2003 ............................................................ 100

COTTON YIELD REGRESSION COEFFICIENTS IN NORTHERN
MOZAMBIQUE, 2003. .......................................................................... 124

TABLE DEFINITION OF THE VILLAGE DUMMY VARIABLES... 126

REGRESSION COEFFICIENTS FOR VILLAGE DUMMY VARIABLES
IN NORTHERN MOZAMBIQUE, 2003. .............................................. 127

QUADRATIC AND TRANSLOG REGRESSIONS FOR MWD TEST13O

REGRESSION COEFFICIENTS FOR BOLLW

ENDOGENEITY TEST .......................................................................... 132
FINANCIAL BT COTTON PARTIAL BUDGET ................................. 139
FOREIGN EXCHANGE PREMIUM CALCULATION, NORTHERN
MOZAMBIQUE, 2003. .......................................................................... 142
ECONOMIC SEED COTTON EXPORT PARITY PRICE, NORTHERN
MOZAMBIQUE, 2003. .......................................................................... 143
FINANCIAL SEED COTTON EXPORT PARITY PRICE, NORTHERN
MOZAMBIQUE, 2003 ........................................................................... 144
ECONOMIC BT COTTON SEED IMPORT PARITY PRICE,
NORTHERN MOZAMBIQUE, 2003. ................................................... 145
PARTIAL BUDGETS TO THE ECONOMY AS WHOLE (ECONOMIC
ANALYSIS), NORTHERN MOZAMBIQUE, 2003. ............................ 146
BT COTTON FINANCIAL ENTERPRISE BUDGETS. ....................... 149
BT COTTON ECONOMIC ENTERPRISE BUDGETS ........................ 154

-x-

FIGURE 5-1

FIGURE 6-1

FIGURE 6-2

FIGURE 6-3

FIGURE 64

LIST OF FIGURES

BT COTTON YIELD AND NET INCREMENTAL BENEFITS ............ 84

FINANCIAL AND ECONOMIC NET INCREMENTAL RETURNS TO
BT COTTON BY SEEDING WEEKS WI I H ONE ADDITIONAL
INSECTICIDE APPLICATION IN NORTHERN

MOZAMBIQUE, 2003. ............................................................................ 93
FINANCIAL AND ECONOMIC NET RETURNS TO BT COTTON BY
SEEDING WEEKS WITH ONE ADDITIONAL INSECTICIDE
APPLICATION IN NORTHERN MOZAMBIQUE, 2003. ..................... 94
FINANCIAL AND ECONOMIC RETURNS TO BT COTTON BY
NUMBER OF INSECTICIDE APPLICATIONS ON SEEDING WEEK
FOUR IN NORTHERN MOZAMBIQUE, 2003 ...................................... 95
FINANCIAL AND ECONOMIC NET INCREMENTAL RETURNS TO

BT COTTON BY NUMBER OF INSECTICIDE APPLICATIONS ON
SEEDING WEEK FOUR IN NORTHERN MOZAMBIQUE, 2003. ...... 96

FIGURE A-l COTTON PRICE PROJECTIONS. ........................................................ 115

FIGURE A-2 COTTON PRICE TREND BETWEEN 1990 AND 2000. ..................... 116

-xi-

GM

JVC

SODAN

CANAM

SANAM

JFS

NGOs

CLUSA

MADER

CIRAD

IFAD

MSU

CINSAM

a.i.

[PM

OLS

BLUE

IPC

ICAC

KEY TO SYMBOLS OR ABBREVIATIONS

Genetically engineered

Joint Venture Companies

Sociedade de Desenvolvimento de Algodao de Namialo
Companhia Algodoeira de Nampula

Sociedade Algodoeira de Namialo

J oao F erreira dos Santos Group

Non-Governmental Organizations

Cooperative League of the USA

Instituto de Algodao de Mocarnbique

Ministry of Agriculture and Rural Development

Centre International de Recherché Agronomic e Developpement
International Fund for Agricultural Development

Michigan State University

Centro de Investigacao e Multiplicacao de semente de algodao de Namilao
Active ingredient

Integrated Pest Management

Ordinary Least Square

Best Linear Unbiased Estimates

Consumer Price Index

International Cotton Advisor Committee

-xii-

MO/DPL Monsanto and Delta Pineland
CAAS China Academy of Agricultural Sciences

ISAA International Service for the Acquisition of Agri-biotech Applications

-xiii-

CHAPTER ONE

INTRODUCTION

This chapter provides the context, objectives, and organization of this study. We
start by presenting the debate around genetically engineered plants, and we move to
explain how these plants might be able to help improve the current low level of
productivity of the cotton sub-sector. Finally we present the expected outcomes from the

study.

1.1 CONTEXT

Genetically modified (GM) plants are the center of a heated debate on the value of
agricultural biotechnology (Pray and Ma, 2001). Large biotechnology companies such as
Monsanto and the biotechnology industry organizations see agricultural biotechnology as
a tool that can help alleviate serious problems of hunger and excessive pesticide use.

A recent International Fund for Agricultural Development (IFAD) report makes a
strong case that an effective use of this technology can provide a way to lessen poverty in
developing countries (Ismail et al., 2001). The main arguments put forward by advocates
of biotechnology are mainly that higher yields, lower levels of labor and pesticide use for
crops such as cotton, soybean, and corn varieties can be achieved if biotechnology is
adopted (Gianessi and Carpenter, 1999; Femandez-Comejo et al., 1998). On the latter
point, recent experiences from India, China and South Africa do not indicate price
difference between conventional and Bt cotton.

Those opposed to the adoption of Bt cotton have expressed concerns including the

fear of damage to the environment, the development of pest resistance to Bt, reduction of
biodiversity, increased profit for multinational companies and impoverishment of small
farmers in developing countries (Ismail et al., 2001 and Altieri, 2001, and Altieri and
Rosset, 2001). For instance, Altieri (2001) argues that GM crops have limited value for
the less developed world for the following reasons: 1) where adopted, they have resulted
in small yield gains, 2) GM crops pose a threat to crop diversity in developing countries.
and 3) GMOs were introduced by private companies that have intellectual property rights
to the main components of the technology, which may limit farmers’ access to them.
Unfortunately, such institutions, intellectual property rights, do not exist in the majority
of African countries, although Zilberman et a1 (2004) recognize that this limitation can be
overcome. Zilberman et a1. (2004) challenge the second point, demonstrating
conceptually that GM crops can serve to maintain and even enhance crop biodiversity.

Most studies on biotechnology impact assessment have been concentrated in the
US, though a few farm level ex-post evaluations have been reported for China, India,
South Africa, Mexico, and Argentina (Pray and Ma, 2001, and Pocket No.5, 2003).
Clearly, more empirical evidence is required to develop a compelling and balanced
assessment of the potential impact of these crops on poor farmers.

In Africa, commercial production of Bt cotton and maize is only taking place in
South Africa and is in its initial stage (Ismail et al., 2001). It is believed that if the small-
scale, resource poor farmers in developing countries gain the same benefits as suggested
by the studies carried out thus far, they should have higher incomes, less health risk and
live in a less polluted environment. Thus, the purpose of this study is to analyze the ex-

ante economic impacts from adopting Bt cotton in Mozambique. The findings from this

study are also expected to lend insights relevant to other Afi’ican countries with
smallholder production of cotton that are likely to consider the adoption of Bt cotton in

Sub-Saharan Africa.

1.2 OBJECTIVES

The main objective of the study is to evaluate the ex-ante economic impacts of
adoption of Bt cotton in Mozambique. More specifically, the objectives of the study are:
1. to determine the financial benefits of Bt cotton to the farmer, and
2. to determine the economic benefits of Bt cotton to the economy as a whole.

To achieve these objectives, a financial analysis is conducted to assess the
expected net benefit from the adoption of Bt cotton from the farmer’s standpoint. An
economic analysis is also performed to evaluate the benefits that are expected from the
standpoint of the economy as whole. This analysis constitutes an innovation into
evaluating benefits since no study on recent experiences on Bt cotton in developing
countries has provided information on the economic vs. private financial benefits of these
crops; most studies have limited their analysis to financial gross margin assessment. The
economic analysis carried out in this study provides an indication as to whether there is a
basis for public sector investment or whether this should be a purely private sector

activity.

1.3 JUSTIFICATION OF THE STUDY

According to Pray and Ma (2001) and Ismail et a1 (2001), more developing
countries should seriously consider allowing the cultivation of Bt cotton because it offers
an effective way of controlling a serious cotton pest, reducing pesticide use, and
improving the health of farm workers. So, governments should be opened to other
biotechnological advances according to their environmental (especially the resistance to a
bollworrn) and safety standard, so that the farmers could be able to choose what best fits
their farming system. The lack of key information to pursue such an informed policy and
strategy is the major limiting factor in Mozambique. As the popularity of biotechnology
spreads worldwide, Mozambique would be interested in analyzing the benefits of
adopting some of the GM crops such as cotton and maize. Inevitably, this adoption will
depend on social, economic and political factors. The present study will be limited to the
economic factors.

This study focuses on cotton because: 1) it is currently cultivated in more than
200,000 ha involving about 300,000 smallholder households (Oson'o and Tschirley,

2003 ), 2) it is ranked second in merchandise exports (afier shrimps), representing 10.6%
of total exports during 1996-1998, and 3) there is evidence that this crop has a potential
for significantly increasing smallholder per capita incomes. Empirical evidence reveals
that in some cotton growing areas cotton sales ranges from 52% to 84% of the total value
of household income (MAP/MSU/F SP, 1991). This report also conclude that the use of
fertilizer and herbicide led to increases in per capita incomes of more than 100% in

Northern Mozambique, and insecticide constitutes the major cost in cotton production.

-4-

Since Bt cotton is designed to reduce insecticide use, assessing its benefits in
Mozambican conditions is crucial. This study generates two main research outcomes.
First, through yield estimation the study will analyze the factors affecting the production
of cotton in Mozambique and their productivity. Second, pursuing financial and
economic analysis, the study provides net gains to farmers’ equity and to the economy as
a whole, and its policy implications. These outcomes constitute the first attempt to
evaluate ex-ante economic benefits from adopting Bt cotton in Mozambique, and are

expected to contribute as a baseline study for decision making related to Bt cotton.

1.4 ORGANIZATION OF THE STUDY

The present study is composed of seven chapters. The first focuses on
introductory issues such as context of the study, objectives, and justification of the study.
The second chapter summarizes the Mozambican cotton sub-sector, and discusses the
strengths and weaknesses of recent studies on performance of Bt cotton in developing
countries. Chapter three presents the data and theoretical framework used to accomplish
the objectives of the study. Fourth, we discuss the results of econometric models used to
predict conventional and Bt cotton yields. Chapter five gathers yield estimates from a
previous chapter and proceeds with financial analysis from the farmers’ standpoint,
whereas chapter six uses the same yield estimates to estimate Bt cotton’s social
profitability. The last chapter draws conclusions, and synthesizes research and policy

implications of the study as well as its limitations.

CHAPTER TWO

BACKGROUND

This chapter has two main objectives. First, it describes the structure and
performance of cotton production in Mozambique. Second, it reports the performance of
Bt cotton in countries that have adopted it and discusses some limitations of those studies

to be addressed on the present study.

2.1 COTTON SUB-SECTOR IN MOZAMBIQUE

This section describes and discusses the structure and performance of cotton sub-
sector in Mozambique. It identifies the participants and their roles in cotton production in
the study zone. In the performance sub-section, the main cotton varieties and their
performance in cotton fields are presented, and related to yields obtained in neighboring

countries.

2.1.1 STRUCTURE

Cotton production in Mozambique has a somewhat a unique history. Prior to
independence in 1975, all cotton in Mozambique was produced by Portuguese settlers
who mainly employed forced labor from the native people. Traditionally large companies
played a major role in promoting the crop in the country. These companies typically
operated as monopsonies.

Afier independence, Mozambique started to seek well—defined paths for

sustainable agricultural development. The role of cotton in agricultural development

continued to generate intense debate (MAP, 1991). Based on the consensus that the future
of Mozambican agriculture attributed to the smallholder farmer (70-80% of the
population), cotton was viewed by decision makers as an important crop to enhance
smallholder income.

Since the colonial times, cotton production in Mozambique was under concession
system]. The current concession areas were established in the early 1990's as part of a
government strategy to improve agricultural performance. Since then, they have
constituted monopolies. These monopoly rights were granted to large Joint Venture
Companies (J VC’s) between government and foreign investment. Currently the most
active JV C’s in Nampulaz, the study area are Sociedade de Desenvolvimento de Algodao
de Namialo (SODAN), Companhia Algodoeira de Nampula (CANAM), Sociedade
Algodoeira de Namialo (SANAM), and three others called the New Operators (NO’s).

SODAN was established in 1992 as a JV C owned by the Joao F erreira dos Santos
Group (JF S) with 51% of the property and the government with 49%. This company was
established in Meconta, Namapa, Muecate, Memba, and Erati districts of Nampula and in
district of Chiure in Cabo Delgado province (Mabota, 2002).

CANAM was established in 1995 also as a JVC but this time between Portuguese
shareholders and government of 75% and 25%, respectively. This company was assigned
to Mecuburi, Ribaue, Lalaua, and Murrupula disctricts in Nampula province.

SANAM was established in 2000 as a private company and the government has

 

' A concession system is a contractual agreement between the government and a private company which
gives the private company allocation of a cotton production area for a given period. The company agrees to
provide technical assistance to smallholders in that area and in return, it receives exclusive rights to market
the cotton from that area.

2 Nampula is the traditional “cotton belt” but more cotton production is now being carried out outside

no vested interest in it. Recently, SANAM built the most sophisticated factory to process
cotton seed in Monapo District, Nampula province. It was granted a concession area in
2003.

The NO’s also consist of totally private companies which operate without
concession rights. Currently, there are three operators actively engaged in cotton
production in Meconta and Monapo districts of Nampula province. They are viewed as
potential competitors3 by the JVC’s but their scale of operation is currently small enough
that they do not cause major concerns. As we can see, the special feature in cotton sector
in Mozambique is the existence of buying agencies or middlemen. Companies are directly
responsible for input delivery and seedcotton buying from farmers, and they have to cover
extensive areas with poor infrastructure (Osorio and Tschirley, 2003).

Most of the smallholder production is carried out under contract. The
smallholders engaged in cotton production in Mozambique are divided in two distinct
groups. Those who plant parcel of cotton along with other farmers in a adjacent block of
land generally made available by cotton companies, and those who produce cotton on
their own fields with average field size between 0.5ha to 1.0 ha. These are mainly
dispersed fields (Osorio and Tschirley, 2003). The NO’s also provide cash credit or in-
kind credit to be used for consumption (e. g., maize flour and dry cassava) as well as seed
and pesticide on credit. In some cases they provide mechanized land preparation on credit

to help farmers prepare land on time.

 

Nampula, e.g. in Cabo Delgado, Sofala, and Tete.

3 Furthermore, it offers higher prices than the JVC’s with concessions. As result, in many cases farmers
from concession areas deliver their cotton to NO’s instead of JVC’s which provide them with technical
assistance. This is currently a source of major conflict between NO’s and JV C’s. Efforts to resolve this are
still in progress.

A small number of farmers are organized in farmer associations. Some of these
associations are locally organized groups that were promoted by the cotton companies as
a way to reduce their costs of extension services and input distribution. Other associations
are as a result of formal association creation by Non-Govemmental Organizations
(NGOs) such as the Cooperative League of USA (CLUSA). These associations serve as a
collective way to link farmers with JVC’s and NO’s with mutual benefits, such as
reduced extension costs for assistance provider and better opportunity to improve farmers
negotiating power collectively, instead of isolated farmers.

The direct role of the government in the cotton sub-sector is via ‘Instituto de
Algodao de Mocambique (1AM). This institution performs several tasks including
statistical management, classification of lint to be exported, and supervision of cotton
production regulations established by the Ministry of Agriculture and Rural
Development, MADER.

In Mozambique there are seven agrochemical companies in inputs market.
Syngenta is a major company, responsible for 20% of the global pesticide market. This
company was formed when the Swiss company Novartis and Swedish-British
AstraZeneca merged their pesticide and seed interests. The remaining companies are
distributors of chemicals namely: Ain-focus, J oao Ferreira dos Santos (JF S),
Agroquimicos, Novagro, Quinagra and Neoquimica.

Three main cotton varieties are planted in Mozambique. ISA-205, recently
adopted from East Africa developed by “Centre International de Recherche
Agronomique et Developpement ” (CIRAD), is a higher yielding variety (with potential

yield of 2,010 Kg/ha) but more susceptible to pests. It requires about 50% more sprays

and expenditure on protection than the local hairy varieties such as A637-24 and REMU-
40. These are more tolerant to sucking pests such as Jassids. With five sprays and
effective overall management, ISA-205 could easily yield more than a ton; A637-24’s and
REMU-40 levels of productivity are about 600 kg/ha with three sprays (Walker, 2003 and
Rodrigues, 2002).

Cotton in Mozambique has a chronic attack of sucking-insects from the very
beginning of the planting of cotton and bollworrns in the middle stage to the end of plant
growth. The Bollwonns4 are the most abundant and damaging cotton pests in
Mozambique. To control for bollworms, farmers apply Pyrethroid insecticides three to
four times as a norm. Organophosphates are typically used in the first two applications to
control sucking-insect5 attacks.

Motivated by the current situation of cotton sub-sector performance in
Mozambique, Boughton et al.(2002) analyzed cotton sub-sector policies and performance
in Sub-Saharan Africa and, based on lessons from Mozambique and Zambia, suggested
four main actions that could be taken to improve the cotton sub-sector in Mozambique:
(1) in the presence of rural credit, input market failure, and weak contract enforcement,
liberalization of the sub-sector appears to be a solution to deal with high concentration of
ginning sector. This solution requires a full revision of the actual concession system and
Osorio and Tschirley (2003) believe that struggles over this system are likely to continue

at the center of policy and program debate in Mozambique’s cotton sector for some time,

 

‘ The “bollworm complex” in Mozambique consist of a number of species including Heliocoverpa armigia
(American bollworm), Diparopsis castanea (Red bollworm), Earias sp. (Spiny bollworm), Pectinophora
gossypiella (pink bollworm), and Dyscercus Sp. (lint steiner). These pests can heavily affect the yield and
lint quality of cotton.

5 This group consist of Empoasca fascialis (jassids known as leafhoppers), Aphis gossypii (cotton aphids),

-10-

(2) exploitation of the capacity of the private sector to deliver public services without
expecting them to do so free of charge, (3) investment in development and multiplication
of new varieties, improvement of pest management, and updating of the grading system.
At least three companies (LOMACO, AGRIMO, and CNA) have attracted some donor
funding for research, but these efforts have not yet resulted in the release of new cotton
varieties (Osorio and Tschirley, 2003), and (4) achievement of effective partnership
between private and public sectors to build institutional policy environments that
encourage technology renewal. Until now, 1AM has been primarily occupied resolving
conflicts over price setting and contract enforcement. This result that a regulatory
institution like 1AM find it difficult to articulate a policy approach that balances the needs
of farmers and of ginning companies as well as to keep focused on challenge of
increasing productivity and quality throughout the system. This deviation of the main role
of IAM have resulted that cotton companies succeeded several years ago to reduce the
export duty rate from 5% to 2.5%, and now lobby to a complete elimination (Osorio and

Tschirley, 2003).

2.1.2 PERFORMANCE

MADER (2001) and Pitoro et a1. (2002) assessed the performance of the cotton
sub-sector and concluded that farm level productivity of cotton in Mozambique was the
lowest compared to neighboring countries (e. g., Zimbabwe, Zambia, and Tanzania).
These studies reported that during the cropping season 1999/2000, average yields ranged

between 460 Kg/ha and 750 Kg/ha in the most productive agro-ecological zone,

 

Memisr'a tabaci (white fly), and Thrips tabaci (Thrips)

-11-

compared to 800 Kg/ha and 900Kg/ha among small farmers in Zimbabwe. In addition,
the International Cotton Advisory Committee (ICAC, 2000) and (Osorio and Tschirley,
2003) reported that the lint yield per hectare in Mozambique is the lowest in Sub-Saharan
Afiica. Osorio and Tschirley (2003) report that the ginning outtum ratio (GOR), defined
as the proportion of cotton fiber obtained from seed cotton, is approximately 34%. This
GOR is lower compared to 38% in Zambia, 40% in Zimbabwe, and as high as 42% in
West Africa. Ceteris paribus, a 40% GOR in Mozambique would reduce the cost of
production of lint by nearly 20%, with major implications for ginner and potentially,
farmer profitability (Osorio and Tschirley, 2003) (See appendix H).

The combination of low productivity of cotton and the long-term downward trend
of world prices (See the appendix A) results in lower profitability for small-scale farmers,
who constitute 70-80% of the population in Mozambique. This has led Mozambique to
seek alternative ways of increasing productivity in the cotton sub-sector. Bt cotton is one

option that is currently being considered.

2.1.3 BIOSAFETY IN MOZAMBIQUE

The biotechnology research and legislation in Mozambique is still in its infancy
stage. Mozambique after ratifying the Cartagena Protocol on Biosafety (CPB) in 2001,
established the National Biosafety Working Group (NBWG), to co-ordinate the biosafety
activities in Mozambique (Munisse, 2002). This is an inter-institutional and multi-
disciplinary team with the task of setting up proper systems for regulatory and
administrative issues, decision making process (risk assessment and management), and

mechanism for public participation. But, these activities are facing several constraints

-12-

such as inadequate financial resources, lack of personnel with sufficient competence, and
poor infrastructure and equipment. All these limitations result that since the group was
formed in 2001, the only output available from this is a legislation proposal that establish
conditions at which Bt maize can be imported.

Monsanto would be interested in expanding its activities to Mozambique in other
to spread fixed costs, and to clear the way for potential future GM maize seed sales, from
the companies’ long-term perspective, but the slow regulatory process is holding back
this intention from the GM Company. At present, the production and commercialization

of GM seeds are forbidden by law in Mozambique.

2.2 Bt COTTON

Bollgard® cotton (Bt) was first developed by Monsanto in the late 19803 by
transforming and incorporating the CrylAc gene {Tom a soil microorganism, bacillus
thuriniensis var. kurstakz' (Edge, J et al., 2001). The Bt gene causes the cotton plant to
produce a protein that is toxic to certain species of bollworms, significantly reducing
chemical pest control costs in infested areas (Falck-Zepeda et al., 2000).

The original goal was to provide farmers with more environmentally fiiendly and
efficacious insect control at reduced costs (Edge, J et al., 2001). According to this author,
the targets of Bollgard cotton are major caterpillar pests, including tobacco budworrns
(Heliots virescens Fabricius), bollworm (Helicoverpa zea Boddie), and pink bollworrns
(Pectinophora gossypiella Saunders), which are some of the most damaging insect pests
worldwide. These pests together accounted for an estimated $391 million in cotton losses

and treatment expenses in the US. in 1995, and $699 million in 1996 (Edge et al., 2001).

-13-

Afier receiving regulatory approvals, Bt cotton was launched commercially in the
United of States in 1996, and subsequently in Canada, Argentina, Australia, China,
Mexico, India, and South Africa.

The number of different GM technologies and their applications in the world are
increasing. Statistics show that the area planted to Bt cotton worldwide increased by 11%
(to 4.3 million hectares) between 1999 and 2001 (Cabanilla et al., 2003). According to a
recent evaluation by Cotton International (2003), the area under Bt cotton is expected to
reach half of worldwide cotton production by 2005, and two—thirds by 2008.

The majority of commercial GM releases, however, have been in the U.S.,
Canada, and Argentina. These three countries account for about 99% of the GM crop area
of the World (Ismail et al., 2001).

So far, Bt cotton has demonstrated remarkable control of some bollworm pests,
particularly the tobacco budworrn and pink bollworm. But, since its commercial release,
it has typically required supplementary insecticide control for these two pests (Stewart, 8.
et al., 2001). Cotton Bollwonns have shown themselves more tolerant than tobacco
bollworm to the eryIAc dendotoxin expressed in currently available Bt cultivars. Other
common lepidopteran" pests such as fall arrnywonns, beet armyworms and soybean
loopers are even more tolerant than bollworms to CrylAc. This is a clear indication that
Bt cotton is not 100% effective in controlling these pests as claimed by Bt cotton
advocates.

In the last few years, though there has been an increase in the number of hectares

being planted to GM crops in developing countries (Pray and Ma, 2001), very few studies

-14-

on the farm-level impact of biotechnology have been published about countries outside
North America. The question that arises is whether small-farmers in developing
countries (Africa, Asia, and Latin America) will enjoy the same benefits.

Two positions are taken in this debate. On the one hand, biotechnology companies
often claim that genetically modified organisms (GMO’s) are essential scientific
breakthroughs needed to feed the world, protect the environment, and reduce poverty in
developing countries. On the other hand, some researchers (e. g., Altieri and Rosset,
2001), disagree and argue that most innovations in agricultural biotechnology have been
more profit-driven than need-driven. They also argue that the environmental performance
of transgenic crops needs more tests, not only on the direct effects on target insects or
weeds but also on the indirect effects on other plants and on organism adaptation. They
believe that the target insects are likely to develop resistance against the toxin spread by
GMOs and that this may have unforeseen consequences. Below, we briefly discuss the
available studies that focus on Bt cotton adoption by small-farmers in developing

countries.

2.2.1 IMPACT OF Bt COTTON IN DEVELOPING AND DEVELOPED
COUNTRIES

This section provides a literature review on Bt cotton performance around the
world and identifies gaps in information that this study expect to complement. Table 2-1

summarizes key parameters of recent experiences in countries that adopted Bt cotton.

 

° A class of bollworms.

-15-

Table 2-1 Yield and Cost advantages of Bt over non-Bt cotton in selected countries.

 

 

 

 

Average
Incremental Net .
Re 0 f B t Yreld Advantage Cost Advantage
COMM/Year cotton
0
/° Non- Difference Number Of Reduction in
gain B t sprays
°"°’ Y' 1d Bt N N or C t
(S/ha) re . 0 on- r. os
N1??- Yreld Kg/ha /o B t Bt sprays (Slha)
India
1998-1999 na na 1501 802 na 80 4 0 4 45
2000-2001 na na 856 619 237 38 4 l 3 42
China (2001) 470 na 3,371 3,186 185 6 28 14 13 626
West Afiica7 33 48 946 860 86 10 na na 3 17
South Africa
1998/9, 2000
Small farmers 43 44 576 395 181 46 na na 5.8 23.6
. Page N°“' 23 11 947 832 115 14 na na na 11
Irrigated
. Ifarge 174 na 4,046 3,413 633 19 na na na 28
rrngated
Indonesia
(2001) na na 2,370 1,820 550 30 10 2 9 na
Mexico
1997 na na 1,580 1,540 40 2 5 3 2 154
1998 335 na 1,710 1,420 290 15 5 2 3 139
Argentina
(1999_2001) 65 na 2,110 1,567 543 35 5 3 2 28
Philippines na 1 5-86 Na na Na 60 na na na na
Australia
98/00 2-28 na Na na -40 -3 l 1.2 7.7 4.7 80
United States8 85 na 2,778 2,359 419 20 na na 2.2 39.5

 

Source: Cabanilla et al. (2003), Traxler (2004), and ISAA (2001), Bennet et al. (2001), and Gouse et a1.
(2002)

n.a. — information not available

 

 

7 At 30% yield gain the Bt net incremental returns were $89/ha (129% gain) and at 45% yield gain the net
incremental returns of Bt cotton were $132/ha (191% gain). This is an ex-ante study.
8 Average for two Bt (NuCOTN33B and NuCOTN35B) and two non-Bt varieties (DP5415 and DP5690).

-16-

MEXICO

The cotton area in Mexico was reduced significantly from 200,000 ha in the 19905
to about 80,000 ha in 2000 because of heavy insect infestation. In the Comarca Lagunera
region in north-central Mexico, infestation of pink bollworm, tobacco budworm, and
cotton bollworm in cotton fields reached critical levels. After the introduction of Bt
cotton, significant reduction in insect pests was observed. Bt cotton provided better
control of pink bollworm and budwonn, and partial control of other major insect pests
(Traxler et al., 2001). The average yields in 1997 were 1,5 80 tons/ha for Bt cotton and
1,540 tons/ha for non-Bt (2.6% yield gain) and in 1998 the Bt cotton yield rose to 1,710
tons/ha whereas the non-Bt decreased to 1,420 tons/ha, equivalent to 20.4% (ISAAA,
2001)

Compared to conventional cotton varieties, the use of Bt cotton resulted in
substantial reduction in pesticide cost by more than $100/ha, and higher yields resulting

in net profit of Bt cotton amounting to nearly $335/ha to farmers.

SOUTH AFRICA

Cotton is grown on about 100,000 ha in South Africa and suffers significant
damage from bollworms (Ismail et al., 2001). In the Makathani Flats, small-scale farmers
planted cotton on average between 1.5 and 3.0 ha with some planting up to 10 ha (Pocket
No. 5, 2003). Only 12% of the 4,530 farmers started growing Bt cotton commercially in
the 1998/99 crop season on 900 ha and in 1999/2000 on 2,155 ha (Ismail et al., 2001).

Since then, an increasing number of farmers are choosing to grow Bt cotton due to

 

Yield was converted from lint to raw cotton in Kg per hectare using 0.42 as conversion factor.

-17-

multiple advantages of increased productivity and decreased use of pesticides (Ismail et
al., 2001). In 98/99, 57% of those farmers in Makathani Flats with greater area than tha
had adopted Bt cotton compared to 7% of those with less than 2.5ha. On average, non-
adopters had a cotton area of about 4.0 ha, while the adopters had an average cotton area
of 6.0 to 8.0 ha. In the 1999/2000 season, 65% of the farmers grew Bt varieties. Ismail et
al. (2001) attributes this increase in the number of farmers growing Bt cotton to reduction
in the number of insecticide applications, reduction in the cost of production, increased
health benefits to less pesticide exposure, (although no value is attached to this
statement), and time saved from having pesticide at a nearby outlet, located 20 Km away
from the plains of Makhatini.

During the 1998/99 crop season, the average yield for adopters was 576 Kg/ha and
for the non-adopters it was 395 kg/ha. In the 1999/00 cropping season, however, the
average yield fell to 409 Kg/ha for adopters and 366 Kg/ha for non-adopters. The lower
yields in this season are attributed to excessive rainfall (50% above the average), causing
flooding and delaying planting of cotton, as opposed to the lower than average rain at the
beginning of the 1998/99 season, which had favored the cotton crop. Observe that yields
per hectare decreased with larger farm size (as expected, given more intensive cultivation
on small areas) and the smaller farmers seem to have gained most yield advantage from
Bt cotton.

Ismail et a1. (2001) found that on average, the reduction in pesticide costs for
adopters rose from 13% in 98/99 to 38% in 99/00, resulting in an increase in gross
margins per hectare (without considering savings in labor) of 11% (R80) in 98/99 and

77% (R276) in 99/00 for adopters. Note that producers with the largest areas of cotton

-13-

have lower yields and gross margins for Bt varieties than the conventional one.

A recent paper presented at the 6th International Conference on Global Economic
Analysis, estimates that with 25% adoption rate of the Bt cotton, the cost of pesticides
will decrease by 25%, labor savings will be 8%, seed cost will increase by 110%, and
yield will increase by 18% in the Southern Afiica Region (Elbehri and MacDonald,
2003). Regarding pesticide use, Bennett et al. (2003) provide findings which support
Elbehri and MacDonald (2003) by demonstrating empirically that an additional litre of
insecticide to control bollworm would increase cotton output by 46.5 Kg/ha for non-
adopters and 94.8 Kg/ha for Bt adopters. Yet the study in South Africa concludes that
yield increases are considered less important than pest control cost savings.

The study in Makhathini Flats provides a first impression of Bt cotton
performance at the farm level in Afiica, but a lot still to be said about it:

0 A reasonable number of farmers, 1/4 of the sample, were using more chemicals
than was recommended and 3% were using less. This can indicate two situations.

First, those farmers are not comfortable with the new technology despite

participating in training sessions. Second, under high pest pressure, the amount of

pesticide to be used was higher than expected because Bt cotton does not control
sucking pests such as J assids and Aphids, which are often the most important
pests in subtropical regions. Despite higher than expected pesticide application,
this study did not quantify the health cost related to this fact, although information

about human health risk in Makhathini Flats was available from Rother (2000)

cited by the author.

- All input provision and cotton marketing in the Flats are concentrated in one

-19-

economic agent (Vunisa), resulting in a monopoly situation. As a result, seed

cotton prices paid to farmers may not reflect the opportunity cost. To handle this

problem, economic analysis would be necessary so that estimates of benefits from
the standpoint of the economy as a whole can be obtained.

0 This study does not discuss how and at what level pest resistance to Bt cotton are
addressed,

All these observations suggest that in similar case studies, the expected costs and
benefits should be built in the analysis to better assess the profitability of Bt cotton. This
study has a merit to shed light on Bt cotton adoption in Africa but the author did not
discuss any policy implications for South Africa in particular or to Africa and other

developing countries in general.

WEST AFRICA

A major public policy issue to tackle in West Africa is whether to adopt or not,
and how to introduce Bt cotton in this region. Cabanilla and Sanders (2003), using a
linear programming model, showed significant benefits and more stable income at the
farm level. Depending on the adoption rates and yield advantage of Bt cotton, they found
fairly significant aggregate benefits. Their estimates indicate that the benefits could range
from $7 million to $67 million in Mali ($/ha); $4 million to $41 million in Burkina F aso;
$5 million to $52 million in Benin; $4 million to $ 7 million in Cote I’voire; and, $1
million to $7 million in Senegal. The reduction in insecticide use was found as an added
environmental benefit. This paper ends up recommending that the most practical

approach to introduce Bt cotton in West Africa is by allowing the developer of the variety

-20-

that is currently in market, to locally adapt the technology possibly in collaboration with a
local entity and later market it for profit (“Gene for rent” mechanism). If the cost to
incorporate the gene to the local varieties is higher than to produce generic Bt cotton
varieties, then the partnership between local and GM entity can be compromised if we
consider that GM entities are more profit driven than need driven companies as Altieri
and Rosset (2001) noted. This is not discussed at all in the paper.

This study starts by fixing the yield of conventional cotton at 860 Kg/ha and
assesses the benefits of yield gains of 10%, 30%, and 45%, combined with reduction in
average sprays of 3.5 based on experiences in developing countries. It is fine to
hypothesize yield gain in an ex-ante study like this but, I feel that higher yield gains are
hard to get and the most likely yield gain by Bt cotton seems to be of 11%. Cabanilla and
Sanders (2003) cite a study indicating that even if the conventional sprays are applied,
farmers in West Afiica incur average yield loss of 23% depending upon the degree of
insect incidence. When compared to reported potential loss of 34% if no insecticide
application is made (Cabanilla and Sanders, 2003), this implies that only 11% of the loss
from insects is saved by chemical sprays. In terms of research question, this study is not
different from ours since both tackle the same problem.

The methodological approach followed by Cabanilla and Sanders (2003) differs
from what we use in the sense that while they use the regional data with linear
programming, we use an estimated yield from an econometric model and from there we
estimated the incremental net income. Then, we estimated the yield gain that would allow
100% returns to investment in Bt cotton using a sensitivity analysis, having in mind the

partial budget approach.

-21-

CHINA

At least three million small farmers in China adopted Bt cotton, growing a total of
at least 0.5 million ha in 2000, resulting in multiple benefits. The smaller and lower
income farmers gained almost twice as much income per unit of land from adopting Bt
cotton as larger farmers and those with higher incomes. This study reports an average
yield gain of 10% over Non-Bt cotton. Moreover, over 80% of the 1999 benefits from
adopting Bt cotton varieties went to farmers. The remaining percent went to the seed
companies (Pray and Ma, 2001).

In terms of insecticide use and related poisonings, the Bt adopters reduced
insecticide use by about 15,000 tons or 49 Kg/ha (an average 13 sprays per hectare were
saved as indicated in table 2-1) (Huang et al., 2002), although no value is attached to this.
Farmers and farm workers were relatively less exposed by insecticide, and preliminary
evidence indicates reduced pesticide poisoning as well.

There are also indications of an enhancement of biodiversity by adopting Bt
cotton, confirmed by the local government authorities in Hibei province in 1997, which
found 31 insect species in Bt cotton fields of which 23 were beneficial while non-Bt
fields contained only 14 species of which five were beneficial (Pray and Ma, 2001). This
is due to the fact that Bt cotton allow a reduced spectrum of insecticide application,
meaning beneficiary insects are less affected.

This study contributes to the biotechnology discussion in developing countries by
providing evidence on the farm level impact of Bt cotton, attempting to measure the
economic, income distribution, environmental and health impacts.

Although the approach used in this study is different from that used by Ismail et

-22-

al.(2001) in the Makhathini Flats, the conclusions are consistent in the sense that both

found that a greater share of the benefits are captured by small farmers. Again, this study

concludes that pest control cost savings are more important than yield gain. Some
limitations can be observed in this study, such as:

0 The study does not quantify economic benefits. This would improve the analysis
because as mentioned by the author, China is liberalizing the marketing of cotton
to let different enterprises trade cotton, which implies that at the moment the
market is still not perfectly competitive, suggesting that economic and financial
results could differ.

- The authors identified environmental benefits but did not quantify and include
them in the benefits of Bt cotton.

- The authors ignored the resistance problem suggesting that it was not an important
issue because farmers in China plant Bt cotton in small, scattered plots which
allow alternative hosts for bollworrns (corn and vegetable surrounding the Bt
cotton plots). This is an optimistic point of view, but the reality is that as Bt cotton
cultivation spreads rapidly, all cotton regions will be populated by Bt varieties and
bollworm resistance may quickly increase, reducing the benefits that are
mentioned here. So, to be more conservative, including these costs would improve

the realism of the situation.

ARGENTINA

In Argentina, the adoption of Bt cotton revealed a similar impact on the local

farmers to those observed in India and South Afiica. For example, in the 1999/2000

-23-

growing season, there was an additional benefit over conventional cotton of $65.05/ha,
due to a higher yield, better lint quality, and cost savings $27.55/ha from two fewer
insecticide applications (Elena, 2000). Qaim and deJanvry (2002) reported yield gain of

35% over Non-Bt cotton.

UNITED STATES

Benefits from Bt cotton include an average increase of 20% in yield (Edge et al.,
2001), a reduction of 2.2 insecticide sprays that translated to approximately 850 tons less
insecticide used nationally in 2001, with significant positive implications for the
environment (ISAA, 2003). Overall financial benefits are $50 to $85/ha, of which
farmers capture the bulk of those benefits. The US. farmers, captured 59% of total
benefits compared with 21% for the technology developers (Monsanto), followed by US.
consumers (9%), the rest of the world (6%), and the gerrnplasm supplier, Delta and Pine
Land Company (5%) (F alck-Zepeda et al., 2000). Contrary to what was observed in South
Africa, the main reason for adoption in the US. is increased yield (54% of sample)

followed by cost savings (42% of the sample) (Klotz-Ingram et al., 1999).

INDONESIA
According to ISAA (2001), Bt cotton in 2001 was the first GM crop to be

commercialized in Indonesia. Preliminary information confirms that Bt cotton provides

effective control of the lepidopteran insect pests that cause significant economic loss

(ISAAA, 2001).

-24-

Bt cotton has increased yield by about 30% and requires only 0 to 3 applications
of insecticide compared with 9 to 12 for non-Bt varieties, and has increased farmers’
income due to higher yields and saving on insecticides. Bt cotton in Indonesia has

increased yield by 0.55 ton/ha in 2001(Cabannila et al., 2003).

AUSTRALIA
ICAC (2001) reports that in 2001/02 Australia grew 404,000 hectares of cotton,

with the highest lint yield in the world of 1,658 kg/ha, for a total production of 670,000
tons; it consumed only 15,000 tons and exported 665,000 tons. Cotton production is
highly mechanized and intensively managed with irrigation and inputs including fertilizer
and insecticide. The deployment and impact of transgenic Bt cottons in Australia has

recently been reviewed comprehensively.

The two principal insect pests are Australian budworm (Helicoverpa punctigera),
and the bollworm, H. armigera. Bollwonn has evolved a high degree‘of resistance to the
various classes of insecticides which are employed in an integrated pest management
(IPM) strategy. Other insect pests include thrips, mirids, aphids, and spider mites. Pest
management accounts for 35 to 40% of operational costs and can range from $220/ha to

$550/ha (ISAA, 2001).

Bt cotton was field tested and commercially released in 1996/97. The area of Bt
cotton increased from 30,000 hectares in 1996/97 to 165,000 ha in regulated annual step
increases of 5% up to a maximum of 30% which was reached in 2000/01, and held at that

same level in 2001/02. Bt cotton was introduced with a license fee of $135/ha which was

-25-

later reduced to $85/ha. Experience with Bt cotton in Australia showed that whereas Bt
provides good control of Helicoverpa early in the season, its effectiveness decreases in
late season requiring supplementary sprays; the performance of Bt cotton has also been

found to vary by location.

A four year yield performance assessment indicates that there is no significant
yield effect from using Bt cotton in Australia: the average yield of Bt cotton over the four
year period was 1,560, and 1,600 Kg/ha for non-Bt cotton (ISAA, 2001). There has,
however, been a 40% decrease in the number of sprays for Bt cotton compared with non-
Bt cotton (6.5 compared to 11.2). This reduction appears to have high priority in
Australia since it is a major contributor to a safer environment and also provides a
foundation on which a more sustainable IPM strategy can be built. In the first two years,
insect control costs, which included the technology fee, were actually higher for Bt
cotton. However, in 1998/99 and 1999/00 the net benefit was in favor of Bt cotton at
$50/ha and $40/ha; respectively. This was translated to a break-even Or modest net
incremental benefits ranging from $3/ha in 1998/99 to $28/ha in 1999/00. Despite modest
returns from Bt cotton, farmers have purchased the full quota of seed available each year

because they are convinced of its environmental benefits and that it provides a foundation

for a sustainable IPM strategy.

The advantages of higher densities of beneficial insects and the reduced negative
effects of broad spectrum insecticides are assigned a high “economic value” in Australia.
Again, this is similar to what South Afiican farmers consider the most valuable benefit

from Bt cotton.

-26-

INDIA

Farm trials in India indicate an increase in yield and a significant reduction in
pesticide use from Bt cotton. Based on experimental data, Qaim and Zilberman (2003)
found that Bt cotton varieties produced by Monsanto Company and Mahyco are resistant
to the three species of bollworrns that plague crops in India, considered in previous
studies as responsible for 50% to 60% of crop damages. This study found that average
yields for Bt cotton hybrid (1,501 Kg/ha) were a remarkable 80% greater than their non-
Bt counterparts (same hybrids but without Bt genes, with average yield of 833 Kg/ha),
and 87% greater than the local cotton hybrid (which had average yield of 802 Kg/ha).
According to Qaim and Zilberman (2003), the amount of insecticide applied on Bt plots

was reduced by almost 70%.

While the studies in Mexico, South Africa, and Chin found reduction in
production cost to be the most important factor in adoption of Bt cotton, the India study
(similarly to that of US.) concluded that the yield effect of Bt cotton is the most

important factor in adoption decisions, followed by savings in production costs.

CONCLUSIONS

From this literature review, one can see that Bt cotton performance can vary from
region to region or even in the same region. One consistent finding is the savings in
production costs, especially that accruing from pest control agents. Yield effect shows no
conclusive pattern. For instance, the North American Bt varieties commercialized by

Monsanto in China are considered ill-suited for handpicking, humid Chinese summers. In

-27-

addition, the Bt gene cannot protect the crop against diseases or non-target pests, which
can wipe out profit margins (Glover, 2003).

Moreover, Pemsl et al., (2004) analyzing the studies in detail found a number of
factors that could have predetermined the positive results. In China, for instance, the
sample size for adopters exceeded non-adopters on average by factor of four. This study
found that non-adopters had negative net returns from cotton production. Another factor
is that in general impact studies of Bt cotton assume that the alternative to planting Bt
crops is routine pesticide application, thus leaving aside other possible pest management
methods, as practiced in integrated pest management (IPM). Therefore, the reduction in
pesticide use due to Bt adoption could have been overestimated by such studies.
Furthermore, measurement of farmer’s pesticide use in developing countries is prone to
numerous sources of error (many brands of pesticide and the practice of mixing several
pesticides). The uncertainty in data due to possible climate and ecosystem variability is
another source of error on benefit estimates.

Although all these difficulties, Qaim and Zilberman (2003) based on recent cross-
country experience hypothesize that the yield effects of Bt crops will be high in two
regions of the world: South Asia and Afiica because these regions suffer from high pest
pressure, low availability of chemicals, and low adoption of chemicals. In this context,
one must ask what yield eflect can be expected in areas of Sub-Saharan Africa, such as
Mozambique, where there is high pressure of sucking pests such as Jassids and Aphids

that Bt cotton does not control eflectively? We turn to these issues in the next chapter.

-28-

CHAPTER THREE

THEORETICAL FRAMEWORK AND DATA SOURCES

This chapter presents the framework used to accomplish the objectives mentioned
in Chapter One. We discuss data availability and its implication for model specification to
answer the question posed at the end of Chapter Two. We start by documenting the Bt
cotton yield effect based on the experiences discussed in chapter two, and turn to the

expected Bt yield in Mozambique based on available information.

3.1 APPROACHES

We start by suggesting that a rational farmer will adopt Bt cotton if he/she can earn
a 100% return on his/her investment. Rough calculations suggest that this will require a
yield increase of 10%-15%.9 Our analytical challenge is to determine whether such yield

gains can be expected in Mozambique.

The literature reviewed so far indicates that Bt cotton yield gains range from a
minimum of 10% up to a maximum of 46%. If we consider that the South Afiican
experience may be similar to Mozambique, Shankar and Thirtle (2003) have
demonstrated that at average rate of pesticide use of 2.2 litres/ha, non-Bt cotton attains

36% of potential yield, whereas Bt cotton at rate of 1.1 litre/ha attains 55% of potential

 

9 We will present more detailed calculations later in the thesis. The farmer will pay about $17/ha for seed,
and will save about $9 on pesticides and $4 on hired labor, resulting in a net additional cost (investment) of
$4. At the 2003 farm gate price of $0.11/kg, the farmer will need 36 kg of additional production to recover
this investment, and an additional 36 kg (72 kg/ha total) to earn a 100% return. This is 10-15% of mean
yields. In this approach, cost savings are considered as part of the benefits.

-29-

output. This implies a 53% yield increase (19/36=0.53) from Bt cotton in addition to 1.1

litres/ha savings in pesticide.

In principle, several methodological approaches can be used to address our
research question as described below according to their categories, including analytical
methods, data needs, and yield estimation. For one of these categories, we present and

discuss the potential approaches.

(a) Analytical methods: A potential framework to address our research question is a
partial budget. The partial budget is a balance which measures the positive and negative
effects of a change in the business. The left side of the balance shows the positive effects
on net income including additional income and reduced costs. To counterbalance this
positive effect, the right side of the balance includes reduced income and additional costs
or the negative effects of the proposed change. The net income effect of the proposed
change is calculated by comparing the sum of additional income and reduced costs with
the sum of reduced income and additional costs. The down side of this approach is that it
does not give the net profit of the farm production system at status quo to assess whether
it is profitable or not at the beginning without incorporating changes. To address this
issue, the enterprise budget approach is suitable. Since our research question is whether or
not to grow Bt cotton, with the implicit assumption that for farmers grow conventional
cotton if it is viable (likely profitable), then comparing a viable existent technology to a
new technology makes the partial budget approach suitable. Partial budget analysis need
information about Bt cotton yield gain that need to be measured. Below we discuss

methods that would allow us to obtain this information.

-30-

(b) Data needs: To gather information on potential Bt cotton yield, three potential
methods are considered. First, if we had a simple empirical relationship from ex-post
assessments in another country that relates the number of pesticide applications and Bt
cotton yield, we could use a “benefit transfer” approach to apply this information in
Mozambique. But transferring such results to a specific case of Mozambique would not
be easy because transferring data from one study area to another requires can imply to

imposing strong restrictions or alternatively requires a set of complex adjustments.

Second, we could use experimental data adjusted to farmers’ conditions in order to
predict the shift factor linked to the technology change. However, such experimental data
are not available in Mozambique. Although these approaches both involve data
collection, they are not directly comparable because while the second approach uses raw

data that can be used in statistical estimation, the first approach consists of transferring

processed secondary data that can be adjusted and directly used in the partial budgets.

The third alternative is to use available data on non-Bt cotton production to predict
the expected yield of Bt cotton afier estimating the yield effect of insects that could be

controlled by the Bt gene. The potential yield estimation approaches are described next.

(c) Empirical estimation: As discussed in section 3.5 of this study, the damage
function and standard production function are candidates for empirical estimation. The
damage function approach is one a candidate to estimate the productivity of pesticides,
but data limitations, especially on the level of pest pressure and amount of pesticide used
(active ingredient), force us to discard this approach. Instead, the standard production

function was used to estimate the productivity of pesticide, hence the conventional and Bt

-31-

cotton yields.

3.2 PARTIAL BUDGET

As we stated earlier, the main question of our study is whether farmers will be
better or worse off growing Bt cotton. This is a typical farm management question that
reflects changes proposed by a manager on a farm. Since all other components of the farm
budget that are not affected by the proposed change will remain constant, a complete farm
budget is not needed to determine the profitability of these specific changes in the
operation of the farm. In this case, a farm manager analyzes only those costs and incomes

that change with a proposed business adjustment.

This exercise can be accomplished by using the partial budget, which means that
only the relevant costs and incomes are included in the analysis (Dalsted and Gutierrez,
1992). Our partial budget will contain benefits and costs that are constituted by four
categories as described by Dalsted and Gutierrez (1992): additional income, reduced

costs, reduced income and additional costs.

Benefits: Since Bt cotton will replace conventional cotton, the Bt cotton yield
gain contributes to additional income. The yield gain will be evaluated at the official farm
gate cottonseed price of $0.11 per Kg in the cropping season 2002/03. Since Bt cotton
reduces pesticide costs, this constitutes a benefit. The reduced pesticide costs will be
evaluated by assessing the cost of each saved insecticide application by using Bt cotton
seed. Another benefit from adopting Bt cotton is the reduction of labor used in insecticide

application. For this analysis, labor was evaluated at the actual wage rate, $0.30 per day.

-32-

The additional income and reduced costs have positive effect on net income and their sum

will indicate the total benefits.

Costs: The production of Bt cotton requires the use of refuge strategy and
purchased Bt seed cotton. The refuge strategy cost is evaluated as the yield difference of
the unsprayed cotton lost on 5% of the cotton plot. The Bt seed cotton financial and
economic prices used in this study were drawn from Ismail et a1. (2001) ($16.6 per ha
equivalent to 202 Rands/ha in 1998/99). Since no reduced income factor is foreseen, the

total cost will be constituted only by total additional costs.

Net income: Indicates the effect of the proposed change on net income by
comparing the sum of additional income and reduced costs with the sum of reduced
income and additional costs. If the additional income and reduced costs (benefits) are
greater than the reduced income and additional costs (costs), then production of Bt cotton
will increase net income to farmer compared to the production of conventional cotton.
This analysis will be performed in financial and economic terms. Gittinger (1982) and
Belli et al. (2001) argue that financial and economic costs/benefits differ only when there
is a market distortion caused either by government policy or other causes. The financial
costs/ benefits are those financially observable by any participants in the production
process, whereas the economic benefits are not directly observable, and appear as a way

to correct for market distortions.

So far we have discussed how we will assess whether or not to grow Bt cotton.
But to make such decision, an estimate of Bt yield is needed. The next section discusses

the framework used to generate these estimates.

-33-

3.3 YIELD ESTIMATION AND ASSUMPTIONS

The effect of Bt cotton was modeled as the difference between mean predicted
yield of Bt and mean yield of conventional cotton under actual farmer conditions (Landis,
2003). Bt yield was modeled from the same data set assuming the recommended practice
of six insecticide applications against bollworms. In economic terms, this means that
when comparing conventional cotton (as actually practiced with an average of three
insecticide applications for bollworrns) to Bt cotton, we considered pest control cost for
conventional and no such cost for Bt cotton. Because Bt varieties may not be as resistant
to J assids as the conventional cotton varieties currently in use, we discount this predicted
Bt yield by the equivalent of one insecticide application to control for sucking-insects.
Failure to discount for susceptibility of Bt cotton to J assids would imply: l)
overestimation of the Bt yield effect, or alternatively, 2) that the Bt gene was introduced

into the local hairy varieties, which is not likely to happen in the short- or medium-run.

Note that the Bt yield we estimate in this manner is likely to be the lower bound of
the Bt cotton yield effect. The continual presence of toxin in Bt plants implies continual
effective control of the most serious aspects of the bollworm complex. Replicating this
control in farmer fields with conventional cotton would require that all farmers be well
trained in pest scouting and IPM practices, that they have the time to apply this
knowledge in a timely manner, and that they have all the chemicals they need at the time
they need them. In practice, the data set we use to develop our estimates (see below) was
generated by farmers with little if any training in scouting, and who operated under

constraints on their time and on pesticide availability that lead to sometimes very

-34-

imperfect timing of the applications that were made. It follows that the marginal
productivity of a pesticide application from this data set will be lower than under optimal
practices, and that we will therefore be reaching only a lower bound of the Bt effect.
According to Silvie et al. (2001), accurate timing of spraying results in a 100-200Kg/ha,

an 13% to 25% cotton yield increase in West Afiica.

All other input levels are fixed at their sample mean. Categorical variables were
fixed at their most common level. The calculations were made for the major cotton
producing districts of Monapo and Meconta in Nampula province and Montepuez district
in Cabo Delgado province. The yield effect of Bt cotton was analyzed in two input use

levels (scenarios) described below.

Low-input Use: The cotton companies had two technology practices which were
captured in the data set, high and low input use level. During the time the data was
collected some companies were providing insecticide, fertilizer, herbicide and tractor for
land preparation in a package to some farmers. These inputs were delivered to farmers of
Montepuez District in the province of Cabo Delgado (Strasberg, 1997). In Nampula, the
cotton companies did not provide more than seed and insecticide to farmers. Thus,
farmers were classified as using high-input package if they received herbicide and
fertilizer besides insecticide, and low-input otherwise. Fertilizer and herbicide use were

obtained from the survey data.

The low-input scenario assesses the combined effect of planting date and
insecticide application to control bollworms. The idea of including the planting date is

based on the concerns about the current poor crop management practices expressed by

-35-

cotton researchers in Mozambique, especially late planting. On average, farmers in

Mozambique perform crop management practices such as land preparation, planting,
weeding and spraying, with 30 days of delay, which tends to reduce yield (Norberto,
2003). The reasonable way we found to include those concerns in the analyses is via

planting date.

Under low-input scenario, in order to estimate yields using Cobb-Douglas (log-
linear model), a constant of one was added to inputs with zero use level (fertilizer and

herbicide).

High-input Use“): This scenario compares yields of conventional and Bt cotton
produced under high-input use production systems (where herbicide and fertilizer are also

used beside pesticides), varying seeding week and insecticide applications.

3.4 DATA SOURCES

The primary source of data for the present study is Strasberg (1997). The
Strasberg data provide very detailed coverage on cost of cotton production in the
Northern region of Mozambique, which we considered a good basis for evaluating
technological change in a cost reduction framework and measure of productivity gain.
This survey was conceptualized to collect data among the major stakeholders in the food
and cotton belt (Strasberg, 1997). The central piece of the research was a survey of rural

households, using a stratified random cluster sample according to whether and how the

 

'0 High-input dispersed farmers received each: No tractor use but 50% of them received 89 Kg/ha of
fertilizer (12-24-12), 3.5 liters/ha of herbicide and a variable number of insecticide applications. But,

-36-

family grows cotton. Similarly, understanding the strategy and operating procedures of
JVCs with respect to smallholder within their respective concession areas was
emphasized. This survey consisted of five rounds of interviews to cotton; maize and
manioc farmers between 1994 and 1996, among the different aspects contained in the
data set, there are information on farm budget, including detailed information on input
use. Farmers were asked to provide information on number of insecticide applications,
herbicide, and fertilizer and their application timing, planting date and rainfall, level of

pest attacks that they have observed during the cropping season.

The second source of information consists of a survey of more than 900 cotton
growers in 2000 in Nampula province designed to collect information on cotton
productivity. To complement this information, other sources of information were used
(e. g., National Agricultural survey 2002). Mabota (2002) and Maumbe and Swinton
(2003) provide perceptions of health risks from pesticide spraying and farmer beliefs
regarding the advantages of protective clothing. Information from these two sources was
sufficient to obtain rough lower-bound estimates of the health-related costs of pesticide
spraying.

Despite the richness of data on cotton production in Mozambique, none of these
provided very specific information that was crucial in the present study. Ideally, we
should be able to predict cotton yield considering the knowledge of damage agents, pest
pressure in this study, because it has high value for economic study as argued by
Lichtenberg and Zilberman (1986). Because this information was not provided in the data

sets we just described for this study, the damage function recommended by some of the

 

fertilizer and herbicide had variation despite standard package being delivered to cotton growers.

-37-

literature in pest management was not used in this study.

For the purpose of our study, the specific information on timing, length and
intensity of pest attacks is important. To minimize the data limitation problem on these
information, we conducted a complementary data collection, the third source of
information, consisting of one-month of field work conducted in Mozambique from mid-
July to mid-August 2003 under the supervision of Michigan State University (MSU)
faculty, and counterparts in Department of Policy Analysis (DAP) within the Ministry of
Agriculture and Rural Development (MADER), and in the National Institute for
Agronomical Research (INIA). The goal of this work was to distinguish the major cotton
pests, the typical timing of their attacks in the crop cycle, the specific pesticides used to
control each of these pests, when and how they are applied, and how effective they are, in
practice, in controlling the pests. For this work, a semi-structured questionnaire was
designed to address these issues and the perceptions on expected performance of Bt
cotton in Mozambique (See appendix B), interviewing key informants. The collected
information allowed us to better interpret the Strasberg data set in order to estimate
productivity of different types of pesticides. Knowing the typical timing of pest attacks in
the crop cycle and its application date, we were able to separate the effect of the two
major groups of cotton pests (bollworm and sucking-insect complexes), which was then

used in our standard yield model presented in section 3.6.

Another limitation of the dataset used in this study is the omission of farmers
management skills that could have lead to biased estimates of the productivity of those
inputs (e. g. herbicide, fertilizer, pesticide, and labor). To get around this problem, we

considered lagged maize yield as an instrumental variable, since maize is the most

-33-

popular food crop planted by cotton growers in the study zone and also because it
competes for resources, especially labor, with the production of cotton. We therefore
estimated two yield models, equation 1 and equation 2, which are presented in appendix
C. Equation 1 does not include maize yield, while Equation 2 does include it in an
attempt to control for selection bias. Regression results show that maize yield is not
statistically significant. This suggests that there was no selection bias based on farm
management ability. We remind you that the interpretation of this instrumental variable
has to be made with caution, because the selection bias could be based on land quality

and other factors that the model does not capture.

All these data limitation have implications on the accuracy of the yield estimation
approach, including the fact that it probably establishes a lower bound on the Bt effect
because for instance, if farmer skills are low, then the marginal productivity of
managerial inputs in actual practice would be lower than under optimal practice, resulting

that the Bt effect is lower.

3.5 THEORETICAL EMPIRICAL MODEL

For a long time, agricultural economists have modeled the output of systems
involving damage agents, including pests, using standard production firnctions. By this
we mean production functions that treat all inputs as if they affect yield directly, as
fertilizer and labor do. The idea supporting a different approach — the damage function
approach - is drawn from biological literature on pest control indicating that pesticides

belong to a class of “damage control” inputs. These inputs are different from

-39-

conventional inputs because they affect output only indirectly, by reducing the damage in
case it occurs. So, their productivity depends on the presence of pest in the field. In
contrast, conventional inputs like fertilizer and labor increase output directly, which
makes them well suited to standard production function approaches. The standard

production function was represented as follows:

(3-1) Q= f(Z)

Where Q is output, and is a function of Z, the vector of all inputs including

damage control inputs.

Lichtenberg and Zilberman (1986) introduced a new way of modeling functions
for inputs such as pesticides by recognizing that the final output is a mix of two
components: the potential output and the losses caused by damage agents. These losses
are a function of environmental conditions that dictate the destructive capacity of the
relevant damaging agents, and of damage control agents through the damage reducing
efforts. So, damage can at most be equal to potential output (at maximum destructive
capacity of damage agent) and at least be equal to zero (maximum control capacity of
damage control agent). Then, the abatement fimction G(X) is defined as a proportion of
the destructive capacity of damaging agent eliminated applying X amount of damage
control agent. Hence, this function will have properties of cumulative probability
distribution [0, l] (G(X) =1 meaning complete eradication of the destructive capacity and
G(X) = 0 meaning zero elimination. Recalling that the characterization of actual output is

a combination of potential output and losses as mentioned earlier, then, Q can be

-40-

expressed as:

(3-2) Q=fl(Z)G(X)l

Where, Q is the actual output, the potential pest-free output is given by f(Z. 1 ), f (.)
is concave in Z and G(X), Z is a vector of all other inputs except damage control agents,
X is a vector of damage control agents, pesticide in our case, and G(X) represents the pest
damage function and is increasing in X. Recently, Shankar and Thirtle (2003) and Qaim
and Zilberman (2003) used this approach to analyze the effects of Bt variety adoption on
pesticide productivity in South Africa, and India, respectively, introducing also a dummy

variable for Bt varieties into the damage control specifications.

Based on their ability to capture the biological response of damage to pesticide
applications, the most popular damage function specifications are exponential, logistic,
and Weibull distribution. Shankar and Thirtle (2003) indicate that damage control
productivity is sensitive to functional form (distribution). They found the exponential
form most advantageous because it allows that some output is still produced when no

pesticides are applied, contrary to the other two functional forms.

This new modeling approach suggested by Lichtenberg and Zilberman
demonstrates why the commonly used production functions were theoretically
inappropriate for inputs that protect yield against damage, and why such functions may
commonly lead to overestimation bias of the productivity of damage control inputs even
in large samples. According to Chambers and Lichtenberg (1994), its dual representation

offers ability to recapture pest damage directly from observed data and permits the

-41-

specification of more flexible representations of damage control technology than

currently exists.

Zilberman and Lichtenberg state that in econometric work, the measurability of X
will be limited by data availability. So, failing to measure X it turns out that the actual
output is determined by factors affecting the potential yield, and the production function
becomes standard production function, resulting in overestimation of the marginal
productivity of damage control inputs. But, Carrasco-Tauber and Moffit (1992) testing
the modeling approach suggested by Lichtenberg and Zilberman using different models,
concluded that in practice, pesticide productivity results did not differ significantly across
models (Carrasco-Tauber and Moffitt 1992). In addition, recently, Shankar and Thirtle
(2003) claim that the direction of the bias cannot be established theoreticallyl 1, contrary

to the overestimation bias suggested by Lichtenberg and Zilberman (1986).

Shankar and Thirtle (2003) found recently in South Africa using a Cobb-Douglas
function that the only difference in estimated pesticide productivity using the standard
production function approach and the damage function approach was the significance and
not the magnitude of the pesticide elasticity. These authors found that while the standard
functional form presented statistically insignificant pesticide elasticity, with the new

approach the pesticide elasticity of production was significant.

 

” While one may expect a Cobb-Douglas (or other conventional functional form) to wrongly estimate the
productivity of pesticides, whether the rnisspecification leads to an under or over estimation of pesticide
productivity cannot be determined prior to estimation.

-42-

3.6 COTTON YIELD MODEL

We acknowledge the importance of setting the damage control function to better
estimate the pesticide productivity but the limitation on data forced us to restrict the
model to standard form. Thus, based on data limitation and considering Carasco-Tauber
and Moffitt’s findings, we proceeded hypothesizing that the level of cotton production per

hectare among Northem Mozambican smallholder is characterized as follows:

(3.3) Q=f(Z), where Z: N,I,H,K,V

Where, N represents the natural or agronomic factors (e. g., land quality, infestation
level, etc.), I represents inputs and support services provided by cotton companies (e. g.,
insecticide, fertilizer, herbicide, etc), H represents household resource allocation and
management decisions (e. g. farmers skill), K represents household characteristics, social
and human capital (e.g., age, and education), and Vis a vector of village dummy

variables.

The rationale for and description of variables used to implement the theoretical

model is discussed below by each category.

NATURAL FACTORS (N)

Land quality, rainfall and levels of pest infestation are the three major natural
factors that are believed to affect cotton productivity. For land quality, the fertility level

of fields was proxied with four-scale categorical land quality variable from (Strasberg,

-43-

1997) based on farmers’ perception of the fertility their land. Under this approach, 61% of
cotton fields were classified as extremely fertile, 32% as fertile, 6% as fairly fertile, and

1% as non-fertile.

The second natural factor is precipitation. Rainfall quantity and its distribution
within the cropping season is essential to agricultural production, since irrigation is not
common in cotton production in the study zone. In the absence of plot-level rainfall data,
to proxy for abnormal precipitation on the farmers plots, farmers were asked to compare
actual quantity and distribution of rain during the production season to a normal year”.
This approach was used based on Strasberg (1997) data set to develop a dichotomous
variable to control for the effect of rainfall on the plot-level production, where a value of
zero was attributed to the cases with implicit normal precipitation and a value of one for

excessive or insufficient rainfall cases.

The third natural factor hypothesized to influence agricultural production is the
level of insect infestation observed during the cropping season. To proxy for insect
infestation levels, a dichotomous variable was computed, where a value of zero was
attributed to farmers that observed infestation levels insufficient to affect yield and a
value of one to those farmers that had seen infestation levels high enough to have affected

their yields.

 

‘2 It is important to remember that the subjectivity of this question can affect the usage of the rainfall
variable, since it depends much in the farrners’ judgment. Normal season can be understood differently
among farmers, but most of them considered a normal year the one that observed lower infestation level,
and higher yield.

INPUT USE (I)

Planting daterEarly planting dates are hypothesized to have a significant effect
on cotton yields. The reason for this hypothesis is that the “Centro de Investigacao e
Multiplicacao de Semente de Algodao de Namialo” (CINSAN) has reported a significant
positive yield effect on cotton yield (Strasberg, 1997) from planting prior to November
15‘“. In addition, field trials indicate that planting after December 30th results in yield a

decrease of about 100 Kg/ha per week (Rodrigues, 2002).

Surveyed farmers in 1995 by Strasberg (1997) reported the timing of each their
crop management activities. With this information, we were able to compute an
explanatory variable, seeding week as a proxy for seedng dates. For seeding week, an
eight-scale variable was computed, where a value of one corresponds to planting during
the second week of November, two for planting which occurred on the third week of

November and so on until seed week eight.

Weeding labor: Weeding between two and ten weeks afier germination is the
most critical period of labor demand for cotton (Strasberg, 1997 and Rodrigues, 2002).
Since the survey provides information on labor used in each activity, we calculated

directly the total weeding labor as total adult-equivalentl3 weeding labor days per hectare.

Insecticide use: Insecticide is the main chemical input used by cotton farmers in
cotton production. The cotton yield model uses the number of insecticide applications on
each field, instead of the quantity of insecticides active ingredient applied because the

survey does not provide data on this.

 

'3 Labor adult-equivalent coefficients are: individuals 7-8 years old =O.3, 9-12 = 0.5, 13-15 = 0.7, males 15-

-45-

According to Rodrigues (2002), farmers are recommended to apply insecticide to
control bollworm starting eight weeks afier germination, and to control for sucking-
insects right after germination up to eight weeks. Combining information on seeding and
insecticide application dates, we were able to separate the insecticide applications into
two major groups according to their timing of application, implicitly equivalent to their

effect in controlling bollworms and sucking-insects.

These two groups were developed to capture the effect of the two major cotton
pests as mentioned before. Assuming that Bt cotton controls bollworm effectively, we
captured the effect of Bt cotton by assuming it to be equivalent to the bollworm control
provided by six insecticide applications on conventional cotton (i.e., six applications
starting eight weeks after germination). In other words, Bt cotton is attributed the
equivalent of three insecticide applications beyond current practice in conventional

cofion.

HOUSEHOLD-CHARACTERISTICS (K)

Household characteristics such as education, age and information stock (access to
extension service) were hypothesized to have a significant effect on cotton yield.
Education (as number of years of formal schooling of the household head) and age of the
head of household were constructed as continuous variables and the correlation
coefficient between them is negative. We expected to see a positive relationship of both

education and age (sometimes representing experience) with cotton yield. Information

 

54 =1 .0, females lS-54=0.85 and individuals >55 = 0.7.

-46-

stock was instrumented as a dichotomous variable, where a value of one was attributed to
the household that had access to extension services from the Cotton Company or other

source and zero otherwise.

VILLAGE-LEVEL VARIABLES (V)

Village-level dichotomous variables (VII-1,. . ., VII-16 are included in the yield
model to control for spatial variation in infrastructure and natural endowment across the
villages. Many infrastructure characteristics may affect productivity but the key types
include distance to the cotton company depots, social infrastructures such as road quality,
access to markets (not only for cotton but also for other goods and services), and
availability of water. The identification of each village used in the model is presented in

appendix D.

3.7 CHOICE OF FUNCTIONAL FORM FOR COTTON YIELD MODEL

The relationship expressed by equation (3.3) is assumed to be concave with
respect to inputs, and Cobb-Douglas, Translog, and Quadratic functions are tested as
described below. From these three models, we selected the one that best described our

data without violating the assumption of concavity for theoretical consistency.

As Griffin et al. (1987) stated, determination of the true functional form of a given
relationship is impossible, so the problem is to select the best form for a given task. For
the purpose of this study, the choice was based in two major characteristics described by

Griffin (1987): concavity and goodness-of-fit. We postpone the discussion of each

-47-

characteristic to Chapter Four. The rationale for choosing and description of the

functional forms to be tested are presented in the next section.

3.8 FUNCTIONAL FORMS

The yield model is estimated using a Cobb-Douglas production function as

follows:

ln( Yield )= a0 + ,6, ln( LANDQUAL)+ ,6, ln(RAINPROB)+ HJINSPESTS
(3 .4) + p, ln(TOTWAE)+ p, ln(SEED WEEK)+ p6 ln(BOLL W)+ ,6, ln(SUCKI)

+ ,6, ln(FERTQ)+ p, ln(1‘rlERBQ)+,6’10 ln(AGE)+ flleIL, + -.+ ,BMVILM
Where”:

YIELD = Kg seed cotton production per hectare,

LANDQUAL = proxy for soil quality (1 if the land is fertile, 2 if is fairly fertile, 3 if
is less fertile, and 4 if the soil is infertile) [N],

RAINPROB = 1 if the rainfall observed was higher than in a normal year (Floods)
and 0 otherwise [N],

INSPESTS = 1 if excessive insect infestation reported on the field during
growing season and 0 otherwise [N],

TOTWAE = total adult-equivalent weeding days per hectare, family and non-
family are included, a continuous variable [I],

SEEDWEEK = seeding week, 1 = earliest seeding (2“d week, November), 2 = 3rd
week, November and so on, integer taking values 1 to 8 [I],

BOLLW = number of insecticide applications assumed to control bollworms
(after eight weeks following germination), an integer variable [I],

SUCKI = number of insecticide applications assumed to control sucking-

-43-

insects (during the first eight weeks after germination), an integer

variable [I],

FERTQ = quantity of fertilizer used in each field in Kg per hectare, a
continuous variable [I],

HERBQ = quantity of herbicide used in each field in liters per hectare, a
continuous variable [I],

AGE = Age of the head of the household in years, an integer variable [H],

VII-1 VII-16 = village level dichotomous variables (see identification for each

village in the appendix D) [V]. The summary statistics of the study sample are presented

in table 3-1.

Table 3-1 Summary statistics of the variables used in the regression model, in
Northern Mozambique

 

Variable Mean Standard deviation
YIELD 801 502.76
TOTWAE 49.8 25.91
SEEDWEEK 3 = 18% and 4 = 22%
INSPESTS 1 = 7% and 0 = 93%
FERTQ 21.0 40.56
BOLLW 2.3 1.56
SUCKI 0.6 .919
HERBQ 8.3 15.48
LANDQUAL 1 = 61% and 2 = 32%
AGE 34.7 8.04
RAINPROB l=18% and 0 = 82%

 

Source: Author’s computation based on Strasberg (1997)
—

 

" [.] Indicates variables according to their categories in the conceptual model.

-49-

The simplicity of algebra characterizing this type of production firnction, the fact
that it facilitates the imposition of concavity (Bennet et al. 2003), and the fact that it
enables coefficient estimates to be interpreted as elasticities, seems to be appeal to
economists and agricultural economists alike in production analyses (Debertin, 1992).
Although this functional form poses all these good properties, it also presents
disadvantages, including its failure to represent details of the three-stage neoclassical
production function, and the imposition of unitary elasticity of substitution (Debertin,
1992); moreover, it does not reveal damage function, so all inputs appear to contribute

directly to yield (rather than some reducing damage).

The Cobb-Douglas firnction has convex isoquants, but it has unitary elasticity of
substitution, not allowing for technically independent or competitive inputs, meaning that
the marginal physical product and average physical product decrease monotonically for
all inputs (Beattie and Taylor, 1985). However, it may be a good approximation for a
production process with inputs that are imperfect substitutes. It is also parsimonious in

parameters.

As referred by Wooldridge (2000), linear forms such as the Cobb Douglas impose
a strong restriction regarding the marginal effect of explanatory variables on the
dependent variables, and constant marginal effect, which is not realistic for many

economic relationships that present diminishing marginal returns.

In order to model a variety of economic phenomena, nonlinear firnctions are seen
as alternative to the strong restriction mentioned above. We have reasons to believe that

most of our parameters of interest such as labor, seeding week and insecticide, fertilizer,

-50-

and herbicide uses involve the second derivatives of the production function, implying the
need for quadratic terms in the explanatory variables. The present study will
accommodate this by considering and testing, the translog and quadratic firnctional forms

in addition to the Cobb Douglas form.

The translog firnction was tested in this study because it is a flexible production
firnction and is justified by its increased accuracy and generality (Beattie and Tailor,
1985). This. functional form retains the ability to impose concavity (decreasing marginal
returns for labor, seeding week and insecticide use). This flexible functional form not
only shares the common characteristics of linearity-in-parameters, but in addition to the
usual log linear terms of the Cobb-Douglas, it also allows quadratic terms, including
interactions among the explanatory variables (Lau, 1986). Because our data possess some
dummy variables, those variables were not considered in interactions and quadratic terms.
The Translog model was estimated as:

ln(Yield)= a0+ a,21n(X,)+a, Z X, +14, 2 2 ln( X, )*ln( X ,1)

i=1 2 i=1j=l

(35> , 2. ..
+5437- 2 X. *1n(X,)+/3.. Z X.

d=lj=1

where 0.0 is the intercept term, X, a vector of dummy variables (the pest
infestation, rainfall probability) and Xv the village dummy variables (location effect), X]-
and Xj vectors of all other variables as presented and defined in equation 3.4 (continuous
and integer variables), and (1,, ad, [3,], and ij are unknown parameters. Bit: 0], when i¢j
and ij= Bjd are assumed throughout to maintain consistency with Young’s theorem. In

the equation 3.5, we separated the two dummy variables from village in order to test the

-51-

interactions between those and other variables (integer and continuous variables). For
instance, we had interest on testing the interactions between INSPESTS and BOLLW as
well as SUCKI and INSPESTS. This kind of interactions would not make sense with the
village dummy variables. Therefore, there was a need for separating those two dummies
from the village dummy group and from the group of the rest because the dummies do not

allow logarithm transformation.

The quadratic functional form provides some additional desired characteristics
beside flexibility (linear, quadratic, and interaction in inputs). The ability of this
functional form to handle data sets with zero values confers a distinct property compared
to the previous two, allowing that the output not equal zero when one or all input levels
are zero, which is consistent with plant biology. This functional form is not
asymptotically convergent and is concave in some cases. The application of this type of
functional form here is intended to capture the decreasing marginal return of seeding
week, weeding labor, and insecticide use level since planting too early can depress yield
and so does planting too late; less weeding labor and insecticide use can also depress

yields. This notion of optimal input use can be modeled with quadratic form as:

10 1 10 10 16
(3.6) Yield: a0.» a, 2 X354. 2 Z X,*X,+p, Z X,
i=1 i=1 i=1 d=l

where do is the intercept term, Xv a vector of village dummy variables (16 village
locations), Xj and Xj vectors of all other variables as presented and defined in equation
3.4. The linear summations for Xj in the equation 3.6 go to 10, instead of 8 shown in

equation 3.5 because in the quadratic form we grouped the non-village dummy variables

-52-

with the other variables, in order to allow interactions. (In the equations 3.4 and 3.5, the
linear summations go to 8 because the two dummy variables cannot be transformed into
logarithms). a], [3,, and Bi are unknown parameters. B]: 13,-, when i¢j is assumed

throughout to maintain consistency with Young’s theorem.

-53-

CHAPTER FOUR

EMPIRICAL YIELD MODEL AND YIELD ESTIMATES

This chapter discusses the properties of the tested models, evaluates alternative
specifications for the cotton yield model, lays out the evaluation criteria for model
selection, then uses the selected model along with other information to generate estimates

of Bt cotton yields.

4.1 MODEL PROPERTIES

To ensure good properties of the models, heteroscedasticity was tested using the
Breuch-Pagan/Cook-Weisburg test (BP/CW) and non-normality using Jarque-Bera test
(JB). The BP test consists in hypothesizing that the variance of each observation in the
dataset is some function of the nonstochastic variables, some or all our explanatory
variables. It tests whether the variance is homoscedastic, and assumes that the error term
is normally distributed (Gujarati, 1995). This test has some limitations, such as the
identification of the auxiliary regression, but the reason why it was chosen is because it is
an asymptotic test. The auxiliary regression specification problem was solved by

considering all our explanatory variables as non-stochastic.

The JB test is an asymptotic test and consists of computing the skewness and
kurtosis measures of OLS residuals and uses them to test whether residuals are normally
distributed: the skewness is equal to zero and kurtosis is equal to three. If the p-value of

the computed Chi-square statistics is sufficiently low the hypothesis that the residuals are

-54-

normally distributed is rejected, otherwise is accepted (Gujarati, 1995).

Results indicated both non-normality and heteroscedasticity were present. To
correct for heteroscedasticity problem, the weighted least square regression was
performed (variance of each observation was used as weights). This procedure resulted in
correction for non-normality problem, and then no correction for non-normality was

necessary.

4.2 EVALUATION AND SELECTION OF A YIELD MODEL

Table 4-1 presents results for the alternative models. We use these results, along
with structural characteristics of each model, to evaluate and select among them based on

the criteria: goodness-of-frt, and concavity for theoretical consistency.

Criteria 1: Goodness-of-fit

Having estimated a particular regression, a natural question is how well the
estimated regression line fits the observations. A popular measure, which we use here, is
the goodness-of-fit (R2), which assesses the proportion of variance of the dependent
variable (y) that is explained by the model. Models that have similar R2 are not easily
judged whether one is superior in terms of the fit of the data. To complement this
limitation of goodness-of-fit, we compared those models in terms of their functional form
specification. The model specification tests were conducted for nested models (Cobb-

Douglas and Translog).

_55-

Table 4-1 Estimated production functions in Northern Mozambique, 2003".

 

 

 

Quadratic Translog Cobb-Douglas

Variable16 Coef. S. E p- Coef. S. E p- Coet'. S. E p-value

value value
LANDQUAL -68.2 37.6 0.071 -023 0.097 0.021 -022 0.102 0.031
INSPESTS -326.3 90.3 0.000 -0.36 0.141 0.012 -029 0.149 0.053
(dummy)
TOTWAE 0.43 0.090 0.000
TOTWAE2 0.05 0.01 1 0.000
SEEDWEEK 2.52 0.993 0.012 -039 0.151 0.011
SEEDWEEK2 .4.1 1.34 0.003 -0.88 0.294 0.003
BOLLW 0.83 0.25 0.001 0.15 0.07 0.021
BOLLW2 15.3 3.55 0.000 0.63 0.156 0.000
SUCKI 40.8 26.75 0.129 0.21 0.905 0.020 0.04 0.043 0.331
FERTQ 5.0 2.78 0.073 0.11 0.043 0.012
HERBQ 11.7 2.47 0.000 0.27 0.043 0.000 0.12 0.057 0.031
AGE 6.07 3.06 0.049 0.30 0.165 0.070 0.12 0.173 0.115
Constant 718.1 231.8 0.002 2.63 1.083 . 0.016 4.76 0.589 0.000
Dependent Variable R2 = 0.6038 R2 = 0.5923 R2 =0.5447
is Yield in Kg/ha Adj R2 = 0.5561 Adj R2 = 0.5408 Adj R2 = 0.4925

F(31,183) = 12.66 F (30, 184) = 11.50 F(24,190) = 9.60
Prob > F = 0.0000 Prob > F = 0.0000 Prob > F = 0.0000

 

Source: Author’s computations based on Strasberg (1997).

 

 

 

‘5 These data were collected in 1994/95 survey but adjusted to reflect 2003. The village locations also
explain some portion of yield variation. To keep the table 4-1 compact, we present the village effect in the
appendix D. Only significant variables are reported in this table.

'6 The variables that were reported in theoretical models but are not reported in this table were dropped
from these models because we failed reject the hypothesis that are jointly equal to zero (F-test was used).

-56-

This consisted of testing for multiple linear restrictions (F -test) that the quadratic terms
on Translog were jointly equal to zero. Using the nested test of Cobb-Douglas against
Translog, we observe that the parameter estimates of the additional variables, especially
the quadratic terms from the Translog, are significantly different fi'om zero (table 4-1).
Consequently, the Translog model is superior in representing our data set, with an
adjusted R-square of 0.55 compared to 0.49 of Cobb-Douglas. Since the Translog
appeared to be superior to Cobb-Douglas, it was then compared to quadratic model using

a non-nested test.

For this, the PE test from Mackinnon, White, and Davidson (MWD) was used,
since nonlinear models are being tested and we assumed that the errors are normally
distributed (Gujarati, 1995). This test consists on testing the null hypothesis that the
difference in predicted values of each model is zero. To implement the MWD test, we

proceeded as follows. We set the null and alternative hypothesis as:

H0: Quadratic model: Yield is a function of regressors, the X’s.

H1: Translog model: lnYield is a function of logs of regressors, the logs of X’s.

Step 1: We estimated the quadratic model and obtained predicted values, ho,

Step 2: We estimated the translog model and obtained its predicted values, lnhj,

Step 3: we computed z, = lnhj- Inho,

Step 4: regressed log of Yield on X’s and z, in the translog model. We rejected H1 if the
coefficient 2; is statistically different from zero by the usual t-test, then the translog is not
the true model,

Step 5: we computed 22 = (antilog of lnh-ho),

-57-

Step 6: regressed the Yield on X’s and 22 on quadratic model. We rejected H0 if the
coefficient 22 is statistically different from zero by the usual t-test, then the quadratic is
not the true model. The results are summarized in the appendix E.

From the regression, it resulted that both Z] and 22 are not statistically significant.
Therefore, both models are correct, so they cannot be rejected. To select the model that
better describes our dataset, we relied on goodness-of-fit (R2) comparison. One
methodological concern arises since different functional forms cannot be compared
directly using R2 because the units of measurement between the two models differ. The
reason for this is that the R2 measures the explained proportion of the total variation in
whatever dependent variable is used in the regression, and different functions of
dependent variables will have different amounts of variation to be explained
(Wooldridge, 2000).

To overcome this problem, the box-cox transformation was made for Translog
model to have comparable R2 to the Quadratic model. This consisted on taking the anti
log of the predicted yield from translog model and correlated to the observed yield. The
square of the correlation coefficient is the R2 of translog model equivalent to that of
quadratic. The results from this comparison led us to accept the translog function as the

one that is superior in terms of fitting our data.

Criteria 2: Concavity for theoretical consistency

The concavity property is important in economic studies mainly for two aspects.

On the one hand, it is an accurate description of the production process, because it reflects

-53-

a situation in which output increases at a decreasing rate (diminishing marginal returns)
as the level of inputs increases. This links to the asymptotic convergence property, since
only concave functions can achieve a maximum output level. On the other hand, only if
the function is concave can inputs that maximize profit be computed fi'om first order

conditions (Griffin et al., 1987).

So far we described how each model fits the data. Then we proceed to discuss
their theoretical consistency. Lau (1986) defines theoretical consistency as the ability of
all chosen algebraic functional forms to capture all theoretical properties of that economic
relationship for an appropriate choice of parameters. Though the quadratic and translog
functional forms have the best fit to the data, the coefficients on the squared BOLLW
term is positive, which means that they violate the diminishing marginal returns property

in this variable of interest. The Cobb-Douglas form does not violate this property.

Unfortunately, the U-shape of the quadratic curve suggests exponentially
increasing returns to pesticide use, which is not consistent with theory, and from what is
known, yields do not grow unboundedly with increasing pesticide use. Another
theoretical problem is presented by the translog model where the linear BOLLW term is
negative, which suggests that, with less insecticide applications to control bollworms, the
yield effect is negative. With all these facts, we have reasons to believe the theoretical
inconsistency is serious enough to reject these models. Table 4-2 summarizes the
performance of each functional form regarding the three choice model selection criteria

presented above.

-59-

Table 4-2 Choice criteria for cotton yield model

 

— —_

 

 

 

Criteria Cobb-Douglas Translog Quadratic
Concavrty 1n BOLLW Yes No No
(desrred)

2 .
Goodness-of-frt (R ) In 0.49 0.55 -
log umts

2 .
Goodness-gf-fit (R ) In _ 0.61 0.53
level umts

 

Source: Author’s corrrputations.
_ r

As we can see from the table 4-2, no one functional form performs well in all
selection criteria. To select the preferred yield model, some properties were treated as
more important than others. We treated concavity as the most important property because
it is consistent with biological grth patterns. Although the R2 was used to compare the
how well the functional forms fit the data, it was attributed low importance compared to
the concavity criteria. Based on this approach, we choose Cobb-Douglas as the preferred

model.

A weakness of the Cobb-Douglas model is that it is not asymptotically
convergent, which implies that no maximum yield will be reached, i.e., yields will
increase indefinitely as herbicide and insecticide are used. This is not consistent with
production biological theory indicating that in plant production there is a level of input
mix that results in maximum yield. This implies that the use of these inputs will be
constrained by the farmers' income, not by biological characteristics of production.

Shankar and Thirtle (2003) in South Africa, observed that Bt cotton reduces the marginal

 

'7 Transformed R2 = squared correlation coefficient between anti log of the predicted yield from translog
model and the observed yield.

-60-

 

Shankar and Thirtle (2003) in South Africa, observed that Bt cotton reduces the marginal

productivity of pesticides as pesticide use increases because Bt cotton acts as a natural

substitute.

We acknowledge that Cobb-Douglas functional form probably underestimated the

Bt yield effect due to omitted insecticide application timing that result in the discussed

problem in the quadratic and Translog models. With late insecticide applications,

inadequate pesticides, and lower application rates than the recommended, ceteris paribus,

it is likely to result in lower (possibly negative) pesticide marginal product; hence, lower

cotton yield, and consequently lower Bt cotton yield gain since our Bt cotton yield

estimation approach depends heavily on the productivity of the pesticide. The omission

on insecticide application timing data cannot be tested but the final result on predicted

yield can give insights about the aggregate effect of the omission as summarized in table

4-3.

Table 4-3 Mean observed yield and predicted yield from the three models in Kg/ha in

 

 

Northern Mozambique.
Y' ld

Mean diffrifence 95%

Data Observations (K g/ha) Std. Error with Confidence
Interval
observed

Mean observed yield from -
our data set (S trasberg, 1997) 215 801 34.28794 734 - 869
Mean predicted yield from +8 '
quadratic model 215 809 26.77509 756 - 862
Mean predicted yield from 215 763 30 67088 -38 703 - 824
translog model '
Mean predicted yield from -56
Cobb-Douglas model 215 745 29.03898 688 - 802

 

Source: Author’s computations based on Strasberg (1997), MADER (2002).

-51-

Our preferred model underestimates the average yield by 7.0% compared to the
average observed yield in Nampula province of 801 Kg/ha, while the translog model
underestimates yields by 4.7%. The quadratic model yield prediction is nearly equal to

observed average yield.

To conclude, although the model specified may explain a part of our results, the
environment in which the estimates were generated from can be contributing the most on
the yield estimates that we have. This is due to the fact that since the data were drawn in
actual farmers’ practices, the productivity of pesticide and other inputs may not reflect

their highest levels.

Insecticide endogeneity

One problem with direct estimation of the production is that the inputs are treated
as exogenous data, when actually their levels are decided by farmers during the course of
the growing season (Shankar and Thirtle, 2003). Although the problem applies to all
inputs, we are interested in looking at insecticide. Insecticide in particular, sometimes is
applied in response to pest attacks, in which a correlation is created between insecticide
variable and the error term in the production firnction (Bennett et al., 2003). This problem
is known as endogeneity. Where endogeneity is present, consistent estimates can be
obtained by suitably including a relevant variable as an instrument. Where endogeneity is
not a problem, the least square estimator is more efficient with the original variable than
instrumental variables (Shankar and Thirtle, 2003). Thus, the original variable would be

preferred in the empirical analysis instead of the instrumental variable.

-62-

Accordingly, the endogeneity problem of insecticide was tested using the
Hausman test. This consisted of running a regression of BOLLW on exogenous
explanatory variables of the original model. The residuals from this regression are then
added to the original model as an additional explanatory variable. If the coefficient on the
residual term is insignificant in the new regression, the null hypothesis of no correlation
between the input and the error term (no endogeneity) cannot be rejected. Appendix F
shows that in fact we cannot reject the null in this case, so we include the actual
insecticide applications in our model. This result is consistent with actual practices in
Mozambique where farmers apply insecticide commonly on calendar basis rather than in
response to pest attacks (Chitlhango, 2001) and the same opinion is expressed by Shankar

and Thirtle (2003) in the context of South Afiican farmers.

4.3 INTERPRETATION OF RESULTS

Overall, the model reported in table 4-1 indicates a high degree of explanatory
power indicated by a p-value of 0.000 for F -statistic. The goodness-of-fit is reasonable,
the adjusted R-square is around 0.49 and the explanatory variables are all consistent with

our expectations.

The positive sign of number of insecticide applications to control for bollworms
and the significance level is consistent with Rodrigues (2002). The marginal product of
insecticide applications on average is 87 Kg for bollworm complex, equivalent to
elasticity of 0.15. Tauber and Moffitt (1992) found 0.136 in 1987 and reported 0.055

using Cobb-Douglas function found by Headley in 1963. Shankar and Thirtle (2003)

-63-

found that, compared to no pesticide application, the non-Bt cotton with current average
application rate of 2.2 liters per hectare can recover 36% of the potential yield. The Bt
cotton with an average pesticide use rate of 1.1 liter per hectare, recovered 55% of
potential output. All these pesticide productivities are well within the confidence interval
of our estimated BOLLW coefficient (95% c.i. = [0.023 — 0.286]). The marginal products
found in our study in table 4-4, are lower than that estimated by Strasberg (1997) for

conventional cotton, who estimated 89 Kg under low-input and 154 Kg under high-input.

The marginal value product (MVP) of insecticide application is $9.57/application,
about triple its cost ($3.31/application). Similar to Tauber and Moffitt’s (1992) findings,
our results suggest that pesticide is under used (below their optimal level), in contrast
with a number of studies of productivity of chemical pesticides in agriculture, which have
concluded that pesticide is over used. Our result is not a surprising outcome if we take
into account that the number of insecticide applications that farmers apply is limited by
how much the cotton companies makes available. Moreover, the insecticide applications

may be late or adequate insecticide may have not been available, hence not effective.

The negative sign and significance level on seeding week coefficient (-0.39) is
consistent with CIMSAN’s findings concerning the importance of early seeding as cited
by Strasberg (1997). The marginal product for a one week delay in planting is equal to 94
Kg of yield loss, which corresponds to a financial loss of $10.34 at a farm gate price of
$0.11/Kg. On average, these figures are higher than that estimated by Strasberg (1997),
who found yield losses from one week seeding delayed of 51 Kg under low-input use, and

88 Kg under high-input use.

-64-

The elasticity of weeding labor per adult equivalent day (0.43) is positive and
significant at 5% significance level. Its marginal product is on average 12 Kg per day;
much greater than 1.6 Kg under low-input and 3.1 Kg under high-input estimated by
Strasberg (1997). At farm gate cottonseed price of$0.11/Kg, the marginal value product
is $1.32 per weeding labor-day adult equivalent. Compared with current wage rates paid
for agricultural labor of $ 0.30'8 (Low et al., 2002) and a minimum wage of $1.25 per day
(RM, 2003), we can observe that cotton production does provide incentives for labor
supply because its productivity in cotton is higher than the wage rates paid for this

activity. Again, this means that the weeding labor is under-used.

The cotton production elasticity of fertilizer use is positive (0.11) as hypothesized
and significant at 5%. This figure is in the range between 0.166 found by Headley in 1963
and 0.082 both in Tauber and Moffitt (1992) using Cobb-Douglas function. Its marginal
product is 8 Kg, corresponding to a marginal value product of $0.88/kg of fertilizer.
Compared with its farm-gate price, one can see that its productivity is above its cost
(under used). Thus, farmers’ efficiency can be achieved encouraging its use until the

marginal value product is equal to the marginal factor cost.

The village effect of three locations (N arnacopa in Monapo, Nacimoja and 25 de
Setembro in Montepuez districts) is statistically significant. Although the village
coefficients of the other locations in the zone appear to be insignificant, they are
correlated with most of the other explanatory variables in the model (see appendix D).

This is an expected outcome in Mozambique were rainfall, which was not well-captured

 

-65-

by the model because of data limitation (no information on quantity of rainfall is
available from those locations), is expected to be localized, which can explain part of

yield variation.

The coefficient of herbicide use is positive (0.12) and significant at 5%. Its
marginal product estimates indicate that using this input increases cotton yield by 20 Kgl,
which is equivalent to a marginal value product of $2.20/l at cotton seed farm-gate price.
This figure is almost 26% higher than its cost. Again, similar to insecticide and fertilizer
(although farmers receive this input at subsidized price), they are still not efficient using it
(under-used). The main reason for this may be because farmers do not have much
experience in using herbicide because it is not common, but also the idea that the cotton
companies define the application rate may result in this level of productivity similar to

pesticides.

Table 4-4 Marginal products, marginal value product, and marginal factor cost for Bt
cotton production at mean values of independent variables.

—
Marginal Marginal Value Input Ratio

 

Variables products (Kg) 122733? cost (S/unit) MVP/MFC
Insecticide for bollworms 87 9.57 3.31 2.89
(No. of applications)

Seed Week (No. of weeks) -94 -10.34 - -
Weeding labor (days) 12 1.32 1.25 and 0.30 1.06 and 4.4
Fertilizer (Kgs) 8 0.88 0.38 2.32
Herbicide (litres) 20 2.20 1.74 1.26

 

Source: Author’s computations based on Strasberg (1997).
—

-66-

4.4 COTTON YIELD ESTIMATES

This section presents and discusses the Bt cotton yield estimates based on the
Cobb-Douglas model. The yield estimates presented in table 4-5 were computed as

described in the section 3.3 (for more details see appendix G).

The Bt cotton yield gain was computed as the difference between the mean
predicted yields of Bt and conventional cotton, discounted the Bt cotton yield impact of
one insecticide application to control for sucking-insects. The Bt cotton yield was
hypothesized as equivalent to an effective bollworm control (six insecticide applications),
whereas the conventional yield was estimated under farmers’ conditions (three insecticide

applications on average). All other inputs were fixed at their most common levels.

According to table 4-5, under typical production practice (seeding week four with
one additional insecticide application), the average yield gain from adopting Bt cotton is
about 6% under both input use levels. As the insecticide is more intensively used, the Bt
yield gain decreases to a minimum of 3.4% when three additional insecticide applications
are made. This has an implication regarding adoption of Bt cotton. Since farmers have
weak purchasing power, they would adopt new technology if its yield advantage is
relatively higher because the actual low pesticide application rates lead to higher yield
loss since Mozambique is located in high pest pressure zone. But, our study does not

provide high yield gains.

-67-

 

 

 

 

 

 

 

 

 

 

 

Table 4-5 Estimated Bt cotton yield and Bt yield gain over non-Bt cotton by seeding
week and number of additional insecticide applications, in Kg/ha in
Northern Mozambique, 2003".
Seedingleek
Level of input use Item 3 4 5 6 7 8
No Conv. cotton yield
74 68 637 618 570
additional (Kg/ha) 833 5 3
pesticide Bt cotton Yield (Kg/ha) 901 806 739 689 667 616
application Bt cotton Yield gain (%) 8.2 8.2 8.2 8.2 7.9 8.1
0"" C°“"' °°“°“ yield 870 779 714 665 627 595
additional (Kg/ha)
pesticide Bt cotton Yield (Kg/ha) 922 824 756 705 664 630
111%th flplication Bt cotton Yield gain (%) 6.0 5.8 5.9 6.0 5.9 5.9
USE TWO cm“ °°“°“ Y‘eld 901 806 739 689 649 616
additional (Kg/ha)
pesticide Bt cotton Yield (Kg/ha) 940 841 772 719 677 643
applications Bt cotton Yieldiain (%) 4.3 4.3 4.5 4.4 4.3 4.4
Three cm“ °°“°“ yield 926 828 760 708 667 633
additional (Kg/ha)
pesticide Bt cotton Yield (Kg/ha) 957 856 785 732 689 655
applications Bt cotton Yield gain (%) 3.3 3.4 3.3 3.4 3.3 3.5
N0 cm“ °°“°“ yield 1 546 1 383 1 269 l 182 1 114 1 057
additional (Kg/ha) ’ ’ ’ ’ ’ ’
pesticide Bt cotton Yield (Kg/ha) 1,672 1.496 1,372 1.278 1,204 1,143
application Bt cotton Yield gain (%) 8.2 8.2 8.1 8.1 8.1 8.1
one C‘m' °°"°“ yield 1 616 1 445 1 326 l 235 1 164 1 105
additional (Kg/ha) ’ ’ ’ ’ ’ ’
HIGH- pesticide Bt cotton Yield (Kg/ha) 1,711 1,530 1,404 1,308 1,232 1,170
INPUT application Bt cotton Yield gain (%) 5.9 5.9 5.9 5.9 5.8 5.9
USE Tw° cm” °°"°“ yield 1 672 1 495 1 372 1 278 1 204 l 143
additional (Kg/ha) ’ ’ ’ ’ ’ ’
pesticide Bt cotton Yield (Kg/ha) 1,745 1,561 1,432 1,334 1,257 1,194
applications Bt cotton Yield gain (%) 4.4 4.4 4.4 4.4 4.4 4.5
Three Conv. cotton yield
1,719 1,538 1,410 1,314 1,238 1,176
additional (Kg/ha)
pesticide Bt cotton Yield (Kg/ha) 1,777 1,585 1,458 1,358 1,280 1,215
applications Bt cotton Yield gaifl%) 3.4 3.1 3.4 3.3 3.4 3.3

 

Source: Author’s computations based on Strasberg (1997), MADER (2000), and MADER (2001a).

 

'9 No additional insecticide application = three insecticide applications for conventional
cotton and zero to Bt cotton. One additional insecticide application = four applications

for conventional cotton and one for Bt cotton, etc. Under low-input and high-input production
systems, the Bt cotton yield is 621 Kg/ha (95% c.i. = [583-660]) and 1,133Kg/ha (95% c.i.= [1,063-1,204]),
respectively.

-68-

With lower Bt cotton yield effect estimated in our study compared to what was
observed in other developing countries, then we ask: 1) What is the farmer’s incremental
net profit with the estimated yield gain, and 2) Given risk, weak purchasing power and
actual farmer’s experience in cotton production, we believe that farmer will adopt such
technology only if it results in higher profit in order to recover fillly their investment.
Therefore, we ask what Bt yield, Bt seed price (license fee), and pesticide savings would
allow a representative farmer to repay the forgone income and earn 100% return on

his/her investment in this technology. These issues are addressed in the next two chapters.

Although the predicted Bt cotton yield, 824 Kg/ha, is higher than that obtained by
South Afiican smallholders (5 76 Kg/ha from table 2-1), the predicted Bt yield gains are
extremely lower compared to what recent experiences have reported so far in developing
countries such as China, and South Afiica. In India where bollworms have a high
destructive capacity that is not well controlled in conventional cotton, researchers found
that pest-related yield losses are 50 to 60% (Puri et al.,1999), and Bt cotton achieved yield
advantage of 37-3 8% over non-Bt cotton in 1998/9 and 2000/01 cropping seasons

(Cabanilla et al., 2003, and Cabanilla and Sanders, 2003).

In South Afiica farmers surveyed showed a Bt cotton yield advantage of 46% in
small farms, 14% in large dryland farms, and 19% in large irrigated farms (Ismail et al.,
2001), and an empirical study estimated 77% yield gain for the average farmer (Bennett et
al., 2003). These high yields are attributed primarily to effective control of bollworms
provided by Bt toxins that is active throughout the growing season, irrespective of level
of infestation (Edge et al., 2001). Those two countries have similar pests to Mozambique,

and we expected that the figures suggested by our study would be close to theirs.

-69-

The difference between our estimates and theirs can be explained by many factors.
On the one hand, our econometric model failed to capture the size of pest management
technical and allocative inefficiencies that could explain part of the differences in yield
attained in other countries. As we can observe from experiences in developing countries
that no one of the experiences in table 2-1 farmers apply insecticide three times during the
cropping season on cotton fields. This can mirror two situations (1) the pest pressure is
higher in those countries compared to Mozambique, as a result, the use of Bt cotton gives
the best performance, and (2) if pest pressure is higher, farmers are likely to be well

skilled to control them.

Our yield estimates are more than 50% lower than the experimental data reported
by Rodrigues (2002). For instance, under recommended six insecticide applications,
planting on fourth week the conventional cotton yield estimate is in average 828 Kg/ha.
Rodrigues (2002) reported that under experimental conditions, yields range from 1,585
Kg/ha for A637-24 to 2,010 tha for ISA-205 with six insecticide applications under
perfect scouting plan. Many reasons can explain this difference, but the environment
where the data was collected explains larger part of the difference, although the
estimation may have also contributed. Since our data was drawn from farmer’s conditions
with poor pest management practices, it is likely to result in an underestimated marginal
product of pesticide. Based on Rodrigues’ (2002) yield results, it is likely that the returns
to Bt cotton would be higher, hence attractive to farmers, because the Bt cotton yield gain

is much higher than the 6% yield gain needed for Bt cotton to be attractive to farmer.

Overall, the average, under typical practice, Bt cotton effect is 5.8%. The average

-70-

estimated Bt cotton yield under low-input use is 824 Kg/hazo, and 1,530 Kg/ha under
high-input. These results suggest that the Bt cotton yield gain is not attractive enough to
encourage farmers to adopt it. Since the insecticide reduction is not likely to occur
because Mozambican farmers already have low application rates, the motivation for
adoption would come from large yield gain, which is not observed from our results or if

the cost of insecticide increases dramatically. Chapters Five and Six estimate these levels.

The analyses done in the following chapters are based on the yield estimates
presented in table 4-3. This analysis is based on the assumption that the resistance
capacity of conventional against sucking-insects (especially J assids) is equivalent to one
more insecticide application compared to Bt cotton. If the real resistance is greater than

what we hypothesize here, all this little advantage of Bt cotton will disappear completely.

 

2° The average yield found by Strasberg (1997) in two major producing zones was 574 Kg/ha, and the
average estimated for cropping season 1999/2000 in four major cotton production zones by Pitoro et. al.
(2002) was 602 Kg/ha. The 95% c.i. = [Yield +/- 0.23] estimated in the week four with one additional
insecticide application along the six insecticide applications.

-71-

CHAPTER FIVE

FINANCIAL ANALYSIS

This chapter assesses farmers’ profitability of adopting Bt cotton by estimating the
financial incremental net benefit of adoption. Partial budgeting is used to compare the
effect of Bt cotton production on the net income of the farmer compared to conventional
cotton production. Crop budget information was drawn fi'om Strasberg (1997), Pitoro et

al. (2002), MADER (2000), MADER/MSU/F SP (2003b), and IAM (2003).

5.1 PARTIAL BUDGET

Using the partial budget approach, additional costs and benefits and the difference
among these are calculated, resulting in incremental net benefit. This value indicates
whether the production of Bt cotton will or not increase a farmer’s net income compared

to conventional cotton.

Farm-level partial budgets are presented for high and low-input use cotton
production schemes for conventional and Bt cotton in Northern Mozambique. Since we
are dealing with Strasberg’s data that were collected some time back, prices were inflated
to 2003 levels using the national consumer price index (CPI) to facilitate the use of more

recent complementary data.

To analyze Bt cotton profitability from the farmer's standpoint, two key
profitability indicators are estimated: net incremental return per hectare and net

incremental return per adult equivalent family labor day (ae).

-72-

5.l.l RELEVANT ADDITIONAL COSTS

The relevant additional production costs are seed and refuge. This section explains

how we evaluated these costs.

Bt COTTON SEED

The additional cost of Bt seed to farmers was evaluated at $16.60 per ha at a
seeding rate of 25 Kg/hau. This figure is equivalent to 202 Rands/ha reported by Ismail et
a1. (2001) in South Africa in 1998/99. Because conventional cotton seed is currently
distributed to all cotton growers fiee of charge by cotton companies, the full cost of Bt

seed is incremental.

REFUGE STRATEGY

Resistance to the insecticidal proteins produced by the soil bacterium Bacillus
thuringiensis (Bt) has been documented in more than a dozen species of insects
(Tabashnik et al., 1998a). This situation is raising concerns about environmental hazards
and driving a search for alternatives to decrease the speed at which the resistance is
spreading. For this purpose, two tactics have been recommended by the literature: refuge
and high-dose strategy (Genissel et al., 2003). The refuge strategy consists in establishing

conditions to ensure that a large proportion of the pest population develops on non-

 

2' Note that farmers in Mozambique use seeding rate of about 40 Kg/ha for conventional cotton seed. We
assume 25 Kg/ha because cotton corrrpanies in Mozambique have expressed the intention to reduce the

-73-

transgenic “refuge ” hosts, generally the crop itself. For cotton, it is recommended that 5
to 20% of the area planted should be non-transgenic. Three main structural refuge options

were suggested by EPA (2001):

1) 95:5 external structured unsprayed refuge. Five acres of non-Bt cotton is
planted for every 95 acres of Bt cotton. The refuge must be planted half a linear mile from
the edge of the Bt cotton field and may not be sprayed with any insecticide labeled to
control budworm, cotton bollworm, or pink bollworm. The refuge must be at least 150
feet wide. Given the smaller farm size that characterizes the smallholder cotton

production in Mozambique; this strategy is likely not to be viable;

2) 80:20 external sprayed refilge. Requires that at least 25 acres of non-Bt cotton
are planted as refuge for every 100 acres of Bt cotton. All cotton may be treated with
insecticides to control bollworms, and the refuge must be planted within one linear mile

or preferably half a linear mile from the edge of the Bt cotton field, and

3) 95:5 embedded refuge, which requires that at least five acres of non-Bt cotton
is planted for every 95 acres of Bt cotton. The refuge must be embedded as a contiguous
block within the Bt cotton field. The refirge may or may not be sprayed with any
insecticide labeled to control budworm, cotton bollworm, or pink bollworm. The size of
the block must be at least 150 feet wide. Similar to this refuge strategy is seed mixture.
With seed mixture the Non-Bt and Bt seed are mixed in the same bag before seeding so

that refuge plants are randomly interspersed in rows (Tabashnik et al., 2003).

Given the lack of a strong theoretical or empirical basis favoring any particular

 

actual seeding rate to 25 Kg/ha.

-74-

spatial arrangement, to select the appropriate arrangement will depend on farmer’s
managerial skills and biological factors. What is important is to guarantee that the refuge
is close enough to the Bt cotton field in order to facilitate the pest movement from Bt

cotton to the refuge field (Tabashnik et al., 2003).

Another resistance strategy suggested in the literature is high-dose. The high-dose
strategy consists of “staking” two or more toxins through transgenic technology in the
crop. This strategy not only delays resistance for an extended period but also reduces the
amount of refuge required (Roush, 1998). Such a technique has been used in many new
transgenic cotton crops, including Bollgard H used in the US. Such varieties have
achieved pest resistance as high as 96% against a broader range of insects (Cabanilla and

Sanders, 2003).

Since production of conventional cotton does not require a refuge strategy, the
introduction of Bt will add this cost to the farmer. For the present study, the 95:5
unsprayed embedded refuge strategy was used. The unsprayed refuge was chosen because
Gould (2000) have observed that the frequency of alleles in the sprayed refirge had
increased by about 2%, but in the unsprayed refuge the frequency had decreased by about
12%. The cost of establishing a 5% refuge was evaluated as the yield loss due to refuge as
described by USDA (2001). This cost was incorporated in the output side by considering

only 95% of the expected Bt yield as total output.

-75-

5.1.2 RELEVANT ADDITIONAL BENEFITS

The additional benefits of producing Bt cotton are increased yield and reduced

insecticide use and labor. Their financial values are calculated as follows.

OUTPUT

The additional yield that Bt cotton generates is evaluated at $0.11/Kg. This was
the official price in the study zone during the cropping season 2002/03 of 2,750 Mt/Kg,
converted to US. dollars using the official average exchange rate of J une—2003 of
23,963.00 Mt per d011a1’22. Cotton seed in Mozambique is classified in two main grades
(first and second grade). We assume that all cotton would be of first grade, since the

companies in practice classify nearly all cotton as first grade.

PESTICIDE SAVINGS

The use of Bt cotton in many countries has been associated with a reduction in
insecticide costs. In our study, we evaluated this benefit at $3.31/application saved by
using Bt cotton. Since typical practice in conventional cotton is three applications, Bt
cotton will need none for bollworm control. The Bt cotton savings in insecticide

applications is evaluated at $9.43/ha (95 % of $9.93/ha because refuge is unsprayed).

 

22 The cotton companies paid the official cotton price and the official exchange rate was close to parallel.

-76-

LABOR SAVINGS

The use of Bt cotton saves labor used to apply pesticides. For financial analysis
only hired labor by the family is included since it represents a cost saved23 . An accurate
hired labor evaluation was possible because the survey distinguished within each activity
the amount of family and non-family labor used and the total compensation paid. In our
study, the hired labor used to apply pesticide represents 15% of total hired labor used
under low-input and high-input levels. This summed up to $4.05/ha of total labor savings

at $0.30 per day.

5.2 PARTIAL BUDGETS AND ANALYSIS

This section presents and discusses the profitability of adopting Bt cotton from the
farmer’s standpoint and determines the level that will allow farmers to repay the forgone
income and earn a 100% return to their investment. Table 5-1 presents a summary of
partial cotton budgets, with mean incremental returns to labor and land (see appendix F

for more details).

5.2.1 ANALYSIS OF PARTIAL Bt COTTON BUDGETS

Table 5-1 shows that Bt cotton grown under actual condition and practices, low-

input use, will generate lower returns than conventional cotton; incremental net return is

 

23 Family labor is not included in the budget. Instead, returns are expressed per unit of family labor.

-77-

negative. A similar pattern is observed under high-input level. According to the partial
budget presented in table 5-1 (see appendix H), under low-input use, Bt cotton on seeding
week four with one additional insecticide application24 (typical practice hereafter), the
incremental net returns to land and family labor is expected to be -$2.03/ha and -
$0.04/adult equivalent, respectively. Similarly, under high-input use, the average Bt
incremental net returns to land is -$1.51/ha and the incremental returns to labor is -
$0.03/adult equivalent day although it gives positive incremental returns with no
additional insecticide application. The positive net incremental benefit of Bt cotton under
high-input level occurs only when Bt cotton does not use additional insecticides.
Moreover, it is an unrealistic scenario because Bt cotton although effective in controlling
bollworms, will need additional insecticide applications to control other pests, including

some lepdopterans not controlled by Bt toxins (Wally, 2004).

The general pattern shown by results in table 5-1 is that the net incremental
benefit of Bt cotton decreases with insecticide applications and with later seeding dates
under both input use levels. Two outcomes can be visualized: 1) the impact of late
planting is again highlighted as we observed from yield estimates, and 2) under high-
input use, the combined marginal effect of insecticide, herbicide and fertilizer in Bt
cotton yield does not result in sufficient yield gain to offset the costs that Bt cotton poses

(seed and refuge) even with pesticide cost savings.

Comparing our results with those of South Afiica, we observe that Mozambican

 

24 The minimum number of insecticide applications required to control bollworms on Bt cotton as low as
one but in most cases two mainly for the control for Jassids (Wally, 2004).

-78-

farmers observe lower yield gains and profitability from Bt cotton. Results in table 5-1
indicate that under a low-input regime (typical practice) the net incremental benefit of Bt
cotton is -$2.03/ha, lower compared to $29.00/ha obtained by small farmers in South

Afiica.

In terms of net benefits, Mozambican farmers are likely to observe with three
additional insecticide applications, on seeding week three Bt cotton net return equal to

40% ($39.36/ha) of what South African small farmers get, $97.77/ha.

Mozambican farmers, moreover, are likely to observe lower production cost
under both low-input use ($0.08/Kg) and high-input use ($0.06/Kg) compared to farmers
in South Afiica ($0.15/Kg excluding weeding labor), and Argentina ($0.13/Kg). This low
cost of production suggests that profits could increase substantially if yields can be

improved while controlling costs.

Finally we note that the higher returns in South Afiica are driven by a price of

$0.30/Kg25 (Gouse et al., 2002), nearly three times that received by Mozambican farmers.

 

25 Gouse et al. (2002) indicates that cotton price in South Africa was SAR 2.75 per Kg.

-79-

 

 

 

 

 

 

eod- 85. 23. 85. cod- So- So- a é A85 oes- EE a sea oz on-
42 3o 2: a: on- R- a: 5:8 .s-oo é zoo-e .32 Ba.- 2 mac-eo- oz oozed-Ea
Em- con. mi. own. man. Elm- 3m- 5 5.- a-E Ba- 2 E22 oz oe- beacon
3.3 8.2 Eé 2.x 5% 2.2 4.3- 8:8 .256 8.- Aa-E can a mean- oz _aeoe-eea oo-fi
86. 3.0- 3o- 86. 3o- 3o- 86. 5 ca Moe-8 one. Ea.- o- Bao- oz .oe- aces-Eda
one m2- 36 :8 So one wmd =88 5:8 8.- oee obe- EE 2 nee-o- oz boo-nod
EN. 92. .23. 8N- and. 8a. 2 .m- a 8.- a-Sc es.- 2 Eso- oz .oe- _aeo--eee as; am:
2.: 9:: ?fi 33 3:2 one». 2 .8. 8:8 seoo 8.- Aa-S- so- a a.Eo- oz Baz-
Sd- 3.? 3o- 8.? 8.? 5o- 86. a .8 zea-e soa- EE 9 EB...- oz 8.: - 30..
Re Be as Bo Be Be ooe =88 .28 8.- er one- an o- 25o.- oz eons-Eda coo-oa-
3. -- corn SN- N3- we. -- 8:8: 2:. a ca :58 ess- o- Eno- oz 65 ode-sea 25
8.2 2.: $3 £3 3.3 an. :- 3: eeeoo .2-8 .8 BE e5- 2 men-eo- oz
Se- 3o- 86. Be. So. Se- 85 a E 0E8 cooe- BE a see oz .oe-
Ned e2- oho 2.0 No..- Ee :3 8:8 .seoo 6.- oee oce- Ee o- ”Eso- oz noes-Eda coo-sod
Rd- and. one- woo- mg. no. So. a .8 $58 be: 2 Eno- oz .oe- _aeee-eee oz
2.2 8.2 ems-N 2.2 3.2 :5. do. _ m 8:8 .28 5.- aée Ea- o- meso- oz
a h w m v m cm: :3:— ..e .964
=32 fickfiE-fim E 2:33— ....2 3652:..— ”En:

 

 

.88 6355302 E0582 E doze-Emma 022809: Em £83 @5003 3 Smog REE 365:5 538 um

Tm 2an

-30-

foam"

DPS-or cone-69a eozoo Mmflsz Ea .888 ASE-m fies-6cm c380 Mug Ago: magma-Sm no woman macaw-39:8 $853- ”ooh-om

 

8.? 8.? 8.? 8.? 8.? 2.? 2.? a 5 see one- EE e- E8- oz on-

R- 8.8 8.- oi 8..- 8._ 8.- e88 .28 .8 one .82 EE 9 nee-o- oz asap-d8

8.? GT 8.3 ET 8.? 8.? a? a é 35 e8. 2 Eso- oz oe- coo-sod

5.: 8.8 8.8 3.8 8.2 8.8 add: 880 .28 .8 BE 82 o- 258- oz Hebe-8a eon-F

8.? 8.? 8.? 8.? 8.? 8.? 8.? a co are 88- Es 9 E32 oz 8:-

B- 88 8- as a? .8.- 8- e88 .2-8 5.- EE .82 EE 6. 3.an- oz mecca-Eda

a? 3.? on? 3.....- on? 3.? 8.? a .8 Gee 8a- o- 58- oz .oe- 8688

4.2: on: 8.6 2 .8 8.2. 8.8 8.3 880 .2-8 coo see 82 o- nan- oz ode-=28 63 mm:
8.? 8.? 8.? 8.? 8.? 8.? 8.? 5 8185-8288 a BBB oz 8.: shun”
R- 88 8- -.- 8..- 8.- 82 =88 .28 .8 ea»: .82 Ea.- o- meso- oz deco-a8 oeoeod

of- ME. 3;- 6.7 of- 3.7 :2- a 8 A25 82 2 E8- oz .8:- _aeo-_8a one

8.: 8.: 8.8 8.8 8.8 8.8 88: :88 5:8 8.- a-E 82 o- man- oz

8.8 :8 :8 8.8 8.8 8,- 8.8 a .8 A85 88- .EE e E8- oz oe-

42 88 82 a: 8- 8s a: =88 .28 8.- oee 8%- EE 9 mEso- oz 8:8:an oeo-ood

e: 88 $8 8.8 :.- a? 8.~ a é see 8...- 2 Eco- oz oe- _aeo-_eea oz

8.8 2.8 8.8 3.8 8.: 8.8 8.8- 8:8 .28 .8 see 82 o- mEdo- oz

 

APEOUV

Tm 2an

-31-

The uncertainty on environmental impacts of Bt cotton is crucial in making
conclusions on the impact of Bt cotton in Mozambique economy. According to Maumbe
and Swinton (2003), distorted policies that subsidize pesticides worsen health hazards in
most African countries. If we agree that pesticides have negative effect on farm workers’
health and that health has a positive effect on productivity, then a technology that uses
less pesticide could provide additional health cost savings and productivity increase as
some experiences with Bt cotton have shown. In this direction, using benefit transfer
approach, we consider that at least 45% to 83% of the farmer’s annual pesticide expenses
correspond to farmer’s health cost, as was found in Zimbabwe. This translates to $4.75/ha
and $8.25/ha that Mozambican farmers are likely to save by using Bt cotton. This would
increase the financial incremental net benefit to a maximum of $2.72/ha and $6.22/ha

under low-input use level, turning Bt cotton attractive to farmer”.

5.2.2 SENSITIVITY ANALYSIS FOR FINANCIAL PROFITABILITY

Following Byerlee and Hesse de Polanco (1986), we assume that farmers will
adopt Bt cotton only if it can repay the forgone income and generate a 100% return to the

farmers’ incremental investment (at typical practice), which is considered as the

 

2° Caution should be exercised interpreting the estimated health cost savings because benefit transfer
approach although is increasingly being used by decision makers as a way of estimating environmental
values suitable for use in benefit cost analysis. However, recent studies examining the validity of benefit
transfer of passive use values estimated using contingent valuation have rejected the hypothesis of
convergent validity (Morrison, et al., 2002). Debate on which approach is correct can be found in literature
but what seem to be consensus is that instead of transferring average values the better way is to transfer the
functions. Although most studies seems to suggest that transferring functions is betters than transferring
average values, both are subject to subject to significant margins of error Pearce and Howarth (2000). So,
the intention of introducing the health cost transferred from Zimbabwe was to have an idea of possible

-82-

incremental cost per hectare that farmers incur to produce Bt cotton ($16.60/ha for Bt
seed cost, $10.19/ha for pesticide savings, and $4.05/ha for labor savings = -$2.36/ha of
investment, $2.36/ha for 100% return of investment). Therefore, the sensitivity analysis
carried out here identifies the level of: 1) Bt cotton yield, 2) Bt cotton seed price, and 3)
pesticide savings that will result in the stated level of return to the investment. The

switching values. for this analysis are presented in table 5-2.

Table 5-2 Estimated switching values for a representative farmer to achieve 100%
returns to investment on Bt cotton, Northern Mozambique, 200327

—
Switching value (percent)

 

 

Variable LOW-IN PUT USE HIGH-IN PUT USE
Bt seed price (%) -23 -17

Bt cotton yield (%) 6 3
Insecticide savings (add. number 1.1 0.8

of applications for bollworms)

 

Source: Author’s computations based on Strasberg (1997), MADER (2000), and
MADER (2001a).

Results from table 5-2 indicate that for a representative farmer to earn 100%
return to investment on Bt cotton, the Bt seed price would have to decrease by 23% to

$12.78/ha from $16.60/ha under low-input and by 17% to $13.78/ha under high-input.

Alternatively, Bt cotton would have to allow the farmer to save an extra 1.1 insecticide

 

health cost savings from Bt cotton to farmers’ household.
27 The switching values were estimated based on 6% Bt cotton yield gain, considering the average average

-83-

applications beyond the current three under low-input and 0.8 under high-input use levels.
In other words, to allow farmers to earn 100% return on investment, Bt would have to
save a total of 4.1 and 3.8 insecticide applications to control for bollworms under low and
hi gh-input use levels, respectively. These estimates are close 3.5 insecticide applications
saved in the recent experiences with Bt cotton. Under low-input use, the Bt cotton yield
(including refuge loss) would have to increase by 6% to 826 Kg/ha, whereas under high-
input use it would need to increase by 3% to 1,488 Kg/ha. Figure 5-1 presents an

illustration of yield gain effect on net incremental benefit for Bt cotton.

Figure 5-1 Bt cotton yield and net incremental benefits

35 q Net incremental benefit to generate 100% return

301 I
25~ I

20 J ’

’ - .- LOW-INPUT
—o— HIGH-INPUT
15 «

Bt cotton Yield Gain (%)

104

O
O
o 1 T Y r r 1 y T j j

-4 -2 0 2 4 6 8 10 12 14 16 18 20 22 24 26
Net Incremental Benefits (S/ha)

 

 

 

 

 

conventional cotton yield, average pesticide savings, average seed cost, and average refuge cost under
typical practice.

-84-

CHAPTER SIX

ECONOMIC ANALYSIS

This chapter performs economic analysis with the objective of evaluating the net
contribution of Bt cotton adoption to society’s welfare. The economic analysis is
designed to account for the fact that market prices often do not reflect economic values
due to imperfect market structures or policies which change relative prices (Aller, 1987).
Gittinger (1982) define economic values (shadow prices) as any price that is not market
price, but it carries the connotation that is an estimated of opportunity cost of the good or
service. The financial values are market prices. The financial and economic values are

equal in an economy free of political or market distortions.

6.1 ECONOMIC PARTIAL BUDGET

To address these issues, economic partial budget analysis is used. Once financial
prices for costs and benefits have been determined, their economic values can be
estimated by adjusting the financial prices for market distortions. This process generates
so-called “shadow prices”. Gittinger (1982) defines shadow price broadly as any price
that is not a market price, but its common usage carries the connotation that it is an
estimate of the economic value of the good or service weighted to reflect income

distribution and savings objectives.

In Mozambique, for instance, farmers in Monapo district paid a subsidized

-35-

insecticide cost of $9.93 to treat one hectare of cotton, instead of the full cost of about
$33.13 per hectare ($11.04 per application). The difference is covered by the cotton
companies. Seed, Sprayers, and extension service are also provided free by the
companies. The level of subsidies varies from company to company, but seed and
pesticide application are subsidized to some extent by all companies. A typical thing that
the cotton companies do is to give inputs “free of charge” to farmers, but pay them lower
cotton prices than they would otherwise pay. So, the effect that the monopoly concession
system and government price setting means prices may not equal opportunity cost, so that

these also have to be adjusted.

Converting the prices of internationally traded goods such as cottonseed and
pesticides to local currency using the official exchange rate may not be correct. Because
the official exchange rate may be protected or controlled, and thus, may not reflect the
true economic value of the country’s currency (Belli et al., 2001). When this distortion
exists, border prices need to be converted to domestic currency equivalents using a
shadow exchange rate. To correct for this difference between financial and opportunity

cost of capital, a shadow exchange rate is calculated as:

M(1+tm)+X(1-t,)
(M+X)

 

(6.1) SER= 0ER*(1+ FXP)= OER 1+

Where, SER is the shadow exchange rate, OER is the official exchange rate, FXP is the
percent overvaluation of domestic currency; M is the CIF price of imports in domestic
currency; t," is the average ad valorem import tarrif rate, X is the FOB value of exports,

and tx is the average ad valorem export tax rate. Using this expression, the average

-86-

overvaluation from 1998 to 2002 was 1.18. This figure was used to compute the shadow
exchange rate to calculate the economic price at farm gate of seed cotton, Bt seed, and

pesticide.

6.1.1 RELEVANT ADDITIONAL COSTS

Similar to financial analysis, Bt cotton seed and refuge strategy constitute the
additional cost of producing Bt cotton. Using the decision tree for determining economic
values developed by Gittinger (1982), Bt cotton seed is valued at the import parity price.
The refuge strategy replaces Bt cotton with conventional cotton in a particular area of the
field. Therefore, the cost of this strategy is valuated at the opportunity cost of forgone Bt

cotton yield and is incorporated as a reduction in output value (see appendix 1).

6.1.2 RELEVANT ADDITIONAL BENEFITS

Producing Bt cotton saves pesticides. Like cotton seed, pesticides are traded goods
and involve real resource use; their economic prices should therefore be estimated at
import parity (Gittinger, 1982). However, data limitations prevented us from proceeding
in this way. Instead, we adjusted the Metical (Mozambican currency) financial farm gate
price by removing the subsidies and dividing by SER. This resulted in an estimated
economic cost of $9.34/application28. As we observed in Chapter F our, Bt cotton

production involves yield gain compared to conventional cotton production. Since seed

-37-

cotton is a traded good and is exported, it is evaluated at the export parity price (see
appendix 1 for more details). Data limitation on port farm costs, forced us to use a rule of
Thumb that cotton companies in Mozambique use that, to cover costs and earn a
"reasonable" profit, they can pay farmers not more than 55% of the FOB price (See the
formula used by cotton companies to propose farm gate price in Appendix A). This
approach resulted in a farm gate cotton price of $0.27/Kg, is equivalent to a 245%
increase, from $0.11/Kg. This price suggests that adjusting for market distortions
imposed by government policy; farmers would be able to sell their cotton at higher prices
than they do. But, as result of those distortions, farmers receive inputs at subsidized
prices on credit, and this credit is not completely recovered by the cotton companies, so
they charge the credit losses as well as profit margin from that higher price, which result
in a lower farm gate price of $0.1 1/Kg. Two problems can be visualized by situation:
First, the monopsony rights attributed to firms in environment where the
government does not regulate the firm’s behavior against abusive practices, encourage
those firms to pay lower prices to farmers with the claim that those companies incur
higher costs in providing inputs and services to farmers. As a point of comparison, from
2000-2002, when Mozambican farmers were paid an average price of $0.13/kg, other
countries paid the following averages: Zambia $0.23/Kg; Zimbabwe $0.20/Kg; Tanzania
$0.18/Kg; Uganda $0.19/Kg, Benin $0.27/Kg; Mali $0.25/Kg. To visualize this problem
we refer to example of Dunavant paying something like $0.12/Kg in Tete, and $0.18/Kg

right across the border in Eastern Province of Zambia. Clearly, the concession system has

 

28 This is close to the what cotton companies claim that farmers should be paying: $11.04/application..

-33-

impact on prices paid. Second, the mechanism used by the government of Mozambique to
define minimum price does not reflect the world market cotton prices (Strasberg, 1997),
as consequence, the minimum price policy benefit more JV C’s than cotton growers.
Cotton production, like that of all other crops, involves the use of farmers’
household capital that usually does not enter in their balance sheet. This capital was
evaluated at its opportunity cost, assuming that is equal to 10% of farmers’ cost on
purchased inputs. Since Bt cotton use less pesticides, it implies also labor savings as we
observed in Chapter Five. Under economic analysis, both hired and family labor is
evaluated at their opportunity costs. The family labor used for insecticide application
represents only 6% of the total family labor, equivalent to a cost of $1.03/ha summing to
$5.08/ha the total labor savings with Bt cotton. Table 6-1 summarizes the economic costs

and benefits used in cotton production in Northern Mozambique (see Appendix I).

Table 6-1 F arm-gate economic cost summary in Northern Mozambique, 2003.

 

Item Economic cost (Slunit)
W

Bt cotton seed (Kg) 1.15
BENEFITS

Insecticide ( per application) 9.34

Labor (per adult equivalent day) 0.30
Opportunity cost of farmer’s capital (S/ha) 0.78 to 0.40

Seed cotton (Kg) 0.27

 

Source: Author’s computations based on Strasberg (1997), Howard et a1 (1998),
MADER (2001b), and MADER (2003b).

 

29 Similarly to chapter five, the refuge cost is incorporated in the output value.

-39-

6.1.3 NET ECONOMIC PROFIT FOR Bt COTTON

Table 6-2 summarizes the economic net incremental returns to Bt cotton by
seeding week and number of insecticide applications. Contrary to what we observed in
Chapter Five, the results in table 6-2 show that Bt cotton production will increase the net

income of the society under both input use levels.

Similarly to what we observed in table 5-1, results in table 6-2 suggest that Bt
cotton will be more attractive over conventional cotton with less use of insecticide. This
pattern is due to the fact that with intensive pesticide use, less pesticide savings are
obtained using Bt cotton. These savings are lower to offset the decreasing marginal
returns to pesticides well captured by the estimation model. Shankar and Thirtle (2003)
observed the same pattern with real data on Bt cotton in South Afiica. The relatively
higher incremental benefit when no additional insecticide applications are used is

basically due to the fact that at this level a zero cost of pesticide is imputed.

The positive net incremental benefit on Bt cotton is driven by the insecticide cost
savings of about $29.00/ha, labor savings of $5.08/ha, and seed cotton price of $0.27/Kg.
Since Bt cotton uses less insecticide compared to conventional cotton, removing
insecticide subsidies (economic price) will make Bt cotton more profitable than
conventional cotton. This result turns out to be a benefit for cotton companies, which pay
all input subsidies to farmers, but a disadvantage to farmers, who benefit from these

subsidies.

Overall, the economic net incremental return to land for Bt cotton under typical

-90-

practice is on average $9.61/ha under low-input and $11.30/ha under high-input use. This
results in net returns of $1 16. 14/ha and $228.36/ha under low and high-input uses;

respectively (see Appendix I for indicative economic enterprise budgets).

-91-

 

 

 

 

 

 

 

 

 

 

 

 

 

l l l
J
>956 6.8.60... .8on awn—<2 v5. .898. mutam 82.6% .5200 want-2 Ago: mbsmabm co woman 8.0.5.3388 m...o£:< non—um...”
8.... 8.. 8... 2.8 N... 8.. .m .2 .28. .2... 2 22... oz .2. 2222......
:8. 2.8. 8.8. 8. 8. :88 .888 .28. 22.8 .m .2 .2... 2 22o. oz 222.82.
8.8. 8.: 8. . 8. 1.8. 8.88 8.88 82... .28 .2 .28. .2... 2 22.... oz .2222. 2...:
8.. 8.. 8.8 8... 8.8 82 ..m .2222. .2... 2 22.... oz .2. 2222......
8.8. 3.8. 8.8. 8.8. :88 8.28 .28. 22.8 .m .2 .2... 2 22o. oz 8.2.8....
8.8. 8. .3 8.8. 8.8. 5.8 8.88 22.8 28 .2 .28. .2... 2 2.22 oz .222... 2.... mm:
8.... 8.... 8.: :.: 8.. . 8.. . .m .2 .28. .2... 2 22... oz 8... 22.88.... flunk
8.8. 2. . m. 8...: 8.8. 828 828 .28. 22.8 .m .2 .2... 2 .22... oz 222.82..
8.8. 8...... 8.8. 8...... 8.8 . N 8.88 22.8 .28 .2 .28. .22 2 .22.... oz .2228. .20
8.... 8.... 8.: 8.: 8.... 8.8 .m .2 .28. .2... 2 2...... oz .2. 2222......
8.8. 2.8. 8.8: 8.8. 828 8.28 .28. 22.8 .m .2 .2... 2 .22... oz 222.82.
8.8. 8.8. 8.8. 8...: 8.88 8.88 8...... 28 .2 .28. .2... 2 22o. oz .2222. oz
8... 8... 8... 8... 8.2 82 .m .2 .28. .2... 2 2...... oz .2. 2222......
8.3. 8.8 8.8 8.8 8.8 8.8. .28. 22.8 .m .2 .2... 2 2.22 oz 222.82.
8.8 8...... 8.8 8.8 8...... .0: . 8...... .28 .2 .28. .2... 2 2.2o. oz .2222... 82:.
8... 8.8 8.. 8.8 8... 8.8 .m .2 2.8. .2... 2 .28. oz .2. 222.22....
8.8 8.8 8.8 8.8 8.8. 8.. . m. .28. 22.8 .m .2 .2... 2 22o. oz 8.282.
w . .8 m . .8 8.8 8.8 8.8 8.8. 22.8 .28 .2 .28. .2... 2 228. oz .2222... 8.2 mm:
88 8.8 8.8 88 .88 8.... .m .2 .28. .2... 2 2...... oz .8... 82.2.2... .......z_
8.8 8.: 8.8 8.8 3.... . 8.8. 8.2.22.8 5 .2 .2... 2 22o. oz 8.22.8.2. -30..
8.8 8.8 3.8 8.8 8.8. 8.8. 22.8 .28 .2 .28. .2... 2 22o. oz .228... 2...
8.2 2.... 8.2 8.... 8.... 2.2 .m .2 .28. 2... 2 2...... oz .8... 82.2.2...
8.8 8. ... 8.8 8...... 8.: . :8... .28. 22.8 .m .2 .2... 2 2...... oz 8.2.82.
8.8 8.8 8.8 .28 8.8. 8.8. 22.5 .28 .2 .28. .2... 2 22o. oz .2222... oz
2 8 8 m e n

-92-

38.8 _a=..==o>:eu I .5
Aafiwv :33 9:33. .3 2:50.. .92 2:853.—

85: 8.. :5... he .26.—

.moom .8: 022.88.. E... 28>» mEBom .3 2.328302 E0582 5 cozoo .m E... 72.2.5250 o. 8:32 882.com N6 2an

Figures 6-1 and 6-2, highlights the positive impact of early seeding has on Bt
cotton profitability. Moreover, these figures illustrate that either the net benefit or the
incremental net benefit of Bt cotton under hi gh-input is about twice that under low-input

use level.

Figure 6-1 Financial and Economic Net Incremental Returns to Bt cotton by Seeding
weeks with One additional Insecticide application in Northern
Mozambique, 2003.

14

 

 

 

 

 

 

10 - ~ ~ .Ch I I C 0 - - - o - - - o . I I Q
,3 -°— Economic Net
.5 8 . Incremental
2‘3 Benefit(High-lnput)
E". 6 ~ " 0' Economic Net
:3
3 Incremental
a: 4 Benefit(Low-Input)
a .
g -t- Financial Net
8 Incremental
a 2 ‘ Benefit(High-Input)
+Financial Net
0 I I r 1 . Incremental
1L 4 5 5 7 3 Benefit(Low-lnput)
-2 "-'.---—-.I—--—-

  
 

   

 

Seeding Week

-93-

Figure 6-2 Financial and Economic Net Returns to Bt cotton by Seeding weeks with
One additional Insecticide application in Northern Mozambique, 2003.

 

 

 

 

 

 

300 -
4
250 ~
g 200 4 -°—Bt Economic Net
23 Benefit(High-Input)
m
g - II - Bt Economic Net
32 150 . Benefit(Low-Input)
fl .
g ‘ ~ . . -a- Bt Financial net
:5 4K m ~ . benefit(High-Input)
0 g . Q
a 100 “ . \ . ~ '0 . . . .
m k . . ' - - .0 _ +BtFinanc1al net
‘ w. . . ~ ‘ ‘ - . .9 , _ _ benefit(Low-Input)
uh . - u . 0
§ .h . - ~
0 7 I T I I
3 4 5 6 7 8
Seeding Week

Comparing economic partial budget results with those obtained in Chapter Five,
under typical practice, as shown in figure 6-3, one can observe that the net profit from Bt
cotton for a private farmer is lower than that generated to the economy as a whole in all
scenarios. Why? First, is because with subsidized inputs, the amount of input savings due
to Bt cotton is lower. On contrast, when the subsidy is removed, the economy generates
transfers, which add to the national gain. Transfers represent the value of the market and
policy distortions in an economy. It is the difference between the financial and economic

prices. For instance, with pesticide economic price of $9.34/application, a transfer of

-94-

$6.03/ha over the financial is created, which is a benefit for the society since it is part of
the savings due to Bt cotton. The source of these market distortions is discussed in

section 6.3.

Figure 6-3 Financial and Economic Returns to Bt cotton by number of insecticide
applications on seeding week four in Northern Mozambique, 2003.

 

 

 

 

 

 

 

 

250 1
200 -
7:?
4:
R
j: 150 - .
E -°— Bt Economic Net
{3 Benefit(High-lnput)
M D """""" D I . . .
S """"" o . . . . ' 0' Bt Economic Net
:5. 100 - ----- o Benefit(Low-lnput)
0 ll— ------- i— -------- i- -------- ‘
m d- Bt Financial net
benefit(High-lnput)
50 ~
— Bt Financial net
benefit(Low-Input)
0 T I l
0 1 2 3

Number of insecticide applications

Figure 6-4 illustrates that the net incremental benefit for Bt cotton under high-
input use level, decrease rapidly with insecticide application than under low-input. From
this figure, we observe that while with one additional insecticide the net incremental for

Bt cotton under high-input is greater than that under low-input, with two additional

-95-

insecticide application, the low-input use level presents higher net incremental benefits

than under hi gh-input.

Figure 6-4 Financial and Economic Net Incremental Returns to Bt cotton by number
of Insecticide applications on seeding week four in Northern Mozambique,

 

 

 

 

 

 

 

 

2003.
25 -
20 -
<
A 15 i1 -°-Economic Net
2 Incremental Benefit(I-Iigh-
«:79 Input)
‘5: 10 - - U ' Economic Net
E Incremental Benefit(Low-
5
3 Input).
ad fi' F mancral Net Incremental
g 5 - BenefitOligh-Input)
§ —Financial Net Incremental
a ‘ Benefit(Low-Input)
O
-5 - - - - _ _ ‘
-10 J

Number of insecticide applications

To conclude, Bt cotton with insecticide subsidies totally removed, and Bt seed
price is not higher than $28.78/ha, Bt cotton becomes attractive for society as shown in
figures 6-1 to 6-4. The low-input use systems perform better than high—input use with

intensive use of insecticide.

-96-

6.2 SENSITIVITY ANALYSIS FOR SOCIAL PROFITABILITY

Since the net incremental returns for Bt cotton to the society is positive and more
than 100% return to investment, we were interested in performing the sensitivity analysis
to examine how much costs could increase or benefits decrease and still have the country
break—even in economic terms. The switching values at typical practice are presented in

table 6-3.

Table 6-3 Switching values by input use level, under typical practice in Northern
Mozambique, 2003.

 

Variables of interest LOW-INPUT USE HIGH-INPUT USE

 

Switching values (percent)

Bt seed price 23 40
Bt cotton yield -5 -2

 

Source: Author’s computations based on Strasberg (1997), MADER (2000), MADER
(2001a)

 

For the economy to break- even producing Bt cotton compared to conventional
cotton, the Bt cotton yield would have to decrease by 5% to 808 Kg/ha under low-input
and by 2% to 1,454 Kg/ha under high-input. Regarding the Bt seed price, the maximum
Bt seed price that the society would be willing to pay is $35.40/ha (23% increase) under
low-input and $40.29/ha under high-input. These figures are lower compared to Bt seed

price of $ 50.00/ha (SAR 350) and Technology Fee of $107.00/ha (SAR 750) that

-97-

Monsanto charges to South African farmers as indicated by Wally (2004). This suggests
that Bt cotton production will neither be profitable to farmers nor to the Mozambican
economy if these prices were to be applied in Mozambique. This has a political
implication, which is to the government provide seed at lower cost to the society. How?
Discussion on this issue is beyond the scope of the present research but Everson (2004)
suggest two mechanisms by which modern GMOs can affect productivity in developing
countries, including “Gene for Rent " where the recipient country agrees with a GMO
company to incorporates the trait in local varieties and pay the technology fee, and the
“Transgenic Breeding " mechanism in which National Agricultural Research System
breeders use modern technology to produce GMO traits of economic value in crop
varieties. In a resource-poor country like Mozambique, spending government funds in
research like this has to be well justified. According to Evenrson (2004) and Bennett et
al. (2001), only few developing countries are positioned to pursue such mechanisms

(notably China, India, Brazil, Argentina, South Africa, and Kenya).

From the two situations we just described, one can see that Bt cotton production
in Mozambique is much more sensitive to Bt yield than to the Bt seed price. So, more
efforts would be devoted on yield improvement, seeking for alternatives to improve
cotton productivity and increase cash income to smallholders that also guarantee positive
social profitability than to concentrate on Bt seed price, which require a complex

mechanism as noted by Evenson (2004) and Bennett et al (2001).

-98-

6.3 NET TRANSFERS

Gittinger (1982) defines transfers as representing shifts in claims to goods and
services from one entity to another and do not reflect changes in national income. For
instance, in the case of subsidized seed and pesticides, farmers simply pay lower price,
reducing his cost, and thereby increases his net benefit. In the economic analysis, when
removing these subsidies, the direction of transfer shift is from farmers to cotton
companies or government. The higher cotton price represents transfer of resources from
cotton companies to farmers if this is full paid by cotton companies, but can also
represent a direct subsidy by the government that is added to what farmer receives in the
market. Therefore, removing seed and pesticide subsidies and other types of distortions,
represents a loss but it is a gain to cotton companies or government, whereas removing
distortions in cotton price will represent a gain for farmers but losses for cotton
companies or government. The net transfer is calculated as the difference between the
gains and losses The average net transfers resulting from market distortion adjustments
under low-input is $7.98/ha and under high-input use level is $9.29/ha. These transfers
are basically composed by input subsidies, especially in pesticides, labor savings and
import/export taxes removed from the financial prices resulting in increased revenues.
The table 6-4 summarizes gains, losses, and the resulting net transfers within the

economy.

-99-

Table 6-4 Total net transfers in the national economy with Bt cotton production,

 

 

 

 

 

 

 

2003.
Level of in ut Gains Losses Net Gain
use P Item (S/ha) (S/ha) (SIM)
Revenues 1.04
Pesticide 18 .09
LOW-INPUT Labor 1.03
USE Bt seed cost 12.18
Total 20.16 12.18 7.98
Revenues 2.32
Pesticide 18.09
HIGH-INPUT Labor 1.03
USE Bt seed cost 12.18
Total 21.44 12.18 9.26
Source: Author’s computations based on Strasberg (1997), MADER (2000),
and MADER (2001a).-

 

 

The lower contribution of revenues on gains is also due to yield difference as
resulted from the establishment of refuge to slow down the development of pest
resistance to Bt toxins. Table 6-4 also indicates that the production of Bt cotton results in

an average loss to the economy of $12.18/ha due to higher Bt seed price.

Adopting Bt cotton for benefit of society, has a diverse set of benefits and costs
such as: l) removing the market distortion on cotton price, farmers benefit from having
the possibility of selling their cotton at higher price. They also lose the opportunity of
getting subsidized inputs (especially pesticides), but because Bt cotton uses less
insecticide, farmers end up absorbing those pesticide cost savings using Bt cotton. Note
that only the efficient farmers are likely to benefit from Bt cotton as found by Gouse et a1
(2002) in South Afiica. With the establishment of refuges in cotton fields, the society in

general will be better off in sense that the resistance will be delayed. Now, we ask to what

-100-

extent the society in general would be worried about pest resistance in cotton if the Bt
cotton was not introduced, 2) the JVC’s are likely to pay higher cotton price, but because
most of them have their own ginners, they are likely to observe gains in volume of yield
processed that can reduce variable costs (economies of scale). By introducing Bt cotton,
new input delivery structure is likely to be launched, this can increase their management
cost in short run, but it can be compensated in the long run by volume of cotton
processed. As we observed in table 6-4, pesticide savings are the larger part of the gains,
cotton companies will benefit from removing the subsidies, since they are the ones that
covers these subsidies (70%). The pesticide subsidies amount $4 million in 225,311 ha on
cotton production in Mozambique. Introducing Bt cotton will cost them to establish new
input delivery structure, resulting in increased management cost in short run, but it can be
compensated in the long run by higher volume of cotton processed. The JVC’s are likely
to benefit from removing the subsidies, since they are the ones that cover the whole
subsidies (70%), and 3) the Government is likely to lose on cotton export and input
import duties, but since new business opportunities can be created, the government can
still generate revenues from licensing fees, and value added tax from increased sales of Bt

seed and other inputs.

-101-

CHAPTER SEVEN
SUMMARY, KEY FINDINGS, POLICY IMPLICATION, STUDY

LIMITATIONS, AND RECOMMENDATIONS

7.1 SUMMARY OF OBJECTIVES, FRAMEWORK, AND STUDY
CONTRIBUTION

The objective of this study was to analyze the potential economic impacts from
adopting Bt cotton in Mozambique. More specifically, we analyzed 1) benefits to
farmers, and 2) benefits to the economy as a whole. To address the research objective
outlined above, we first estimated mean conventional and Bt cotton yield using a
regression model, followed by the partial budgets to estimate the net incremental benefits
from adopting Bt cotton for farmers and for society. Given the limitation of the partial
budget approach in providing estimates of net returns, we complemented the analysis by
developing enterprise budgets (See Appendix J).

This study is the first attempt to analyze the potential economic benefits of
introducing Bt cotton in Mozambique. It makes a methodological contribution by
illustrating an approach to conduct an ex-ante analysis with limited information on pest
management practices. This approach can be useful in benefit to other developing
countries with smallholder cotton production, who may want to assess the potential
benefits of Bt cotton where similar data limitations may complicate predicting the

potential economic benefits of Bt cotton. The estimates presented in this study are

-102-

preliminary and should be refined as more information becomes available, especially

regarding factors not included in the analysis.

7.2

SUMMARY OF FINDINGS

The main findings of the study are:

Under current practices, especially low level of insecticide applications, adoption
of Bt cotton is not expected to increase net income for farmers. For Bt cotton to
be attractive to farmers, its yield including refuge loss would need to increase to 826
Kg/ha (6%) and to 1,488 Kg/ha (3%) under low and high-input use, respectively.
Alternatively, the Bt seed price would have to decrease at least by 17%, to $13.78/ha.
These results are driven by the estimated Bt cotton yield effect. The estimated yield
gain fi'om Bt cotton in this study is lower compared to what recent experiences have
reported so far in developing countries such as China, India, and South Africa, which
obtained yield gains of 6% to 80%. The high yield gains obtained in some of those
countries are attributed primarily to effective control of bollworms provided by Bt
toxins, which are active throughout the growing season, irrespective of level of
infestation (Edge, J. et al, 2001). Our estimation approach did not capture this full
effect. Instead, only the lower bound of Bt toxins effect was captured. Lower Bt
cotton yield gain is also partly due to lower pest infestation in Mozambique, where
farmers with only three insecticide applications obtain higher yields than what South

Afiican farmers obtained with higher application rates. If farmers were using

-103-

recommended pest management practices (higher application rates), it is likely that Bt
cotton would be attractive to them because there would be more savings of pesticide

costs.

Many other factors could be affecting the level of yield gain besides bollworm
infestation level and input market failure. Qaim and Zilberman (2003) have
hypothesized high yield effect in countries with high pest pressure and weak input-
markets, such as in Southern Africa and areas of Asia. Mozambique and South Africa
are located in a medium- to high lepidopteran pest infestation zone where
theoretically we would expect to see high to reasonably high yield gain, but the
evidence in this study is not conclusive, although South Afiican experience has

shown higher yield gain.

Mozambican farmers have lower production costs under the low-input use
regime ($0.08/Kg) compared to farmers in South Africa ($0.15/Kg excluding
weeding labor), and Argentina ($0.13/Kg). These lower production costs mean that
profits could increase substantially if yields are improved while controlling costs,

taking into account that all inputs are under-used.

From society’s point of view, Bt cotton appears to increase net income to the
Mozambican economy under both input use levels. Yields could fall as much as

5%, to 808 Kg/ha under low-input and by 2%, to 1,454 Kg/ha under high-input, and

-104-

the country would still break even. The society also will observe gains fiom adopting
Bt cotton with seed cost as high as $3 5/ha and $40/ha under low and high-input,

respectively.

7.3 POLICY IMPLICATIONS

The key research finding of this study is that, if farmers were to pay economic cost
of pesticides and be paid export parity prices for seed cotton, they could pay the economic
cost of Bt seed and still earn more profit with Bt than with conventional cotton. In fact,
due to lower pesticide use in Bt cotton, its advantage over conventional is higher when
farmers pay full pesticide costs, which is more likely to be achieved with more
competition. The need for more competition is driven by the finding that the economic
price of raw cotton is far above what companies are actually paying. This very low price
paid to farmers inhibits the adoption of new cotton technology, including Bt cotton,
which is likely to affect the coordination along the product chain. Boughton et al. (2002)
and Tschirley and Zulu (2003) analyze the key performance dimensions for the cotton
production chain in several Afiican countries, including the level of productivity and
quality achieved throughout the chain and the extent to which it pays a competitive share
of the chain’s total value-added found that there is a need to find an effective balance
between the often conflicting needs for coordination and competition in the sector as
illustrated by the Zambian experience. Based on the fact that Mozambique presents lower
average export value and lowest average producer share of any Southern African (SA)

country (Boughton et al. 2002), the challenge for Mozambican’s cotton sector can be

-105-

conceived as how can it generate a workable level of competition while strengthening the
coordination that is needed to increase export value per hectare. In the presence of lower
productivity as noted earlier by Pitoro et al. (2001) due to inadequate technical assistance
provided by cotton companies and lower seed quality, there is an urgent need for a strong
collaboration between government, private companies, and farmers to build an
institutional policy environment that encourages technology renewal, helps solve endemic
rural credit market failure, and ensures necessary coordination to meet strict quality
requirements of modern spinning and weaving technology (Boughton et al., 2002). This
partnership should facilitate obtaining funds to invest in the main areas of the suggested
public/private partnership such as 1) development and multiplication of new varieties
(including biotechnology research), 2) implementation of pest management, and 3) update

the grading system.

1. Based on the key findings of the study, an issue of concern is how to open the
sector to greater competition while not exacerbating credit repayment problems
taking into account the impact of the concession system in cotton pricing policy in
Mozambique. In principle, the simple policy (liberalization or regulated monopoly) is
not sufficient for cotton sector achieve desired performance in the absence of rural input
and credit markets. Thus, identifying the elements of joint public/private strategies is
necessary to improve cotton sector performance as indicated above. The varying
experiences in Africa suggest that some level of competition does not necessarily worsen

credit repayment as long as there is a favorable environment for innovation in the input

-lO6-

distribution and credit systems. Zambia and, until recently, Zimbabwe have been good
models of relatively concentrated but competitive systems that combined good prices to
farmers with effective input delivery and credit repayment.

Boughton et al. (2002) indicate that Mozambique’s experience shows that a high
level of policy-induced coordination is a necessary but not sufficient condition for
innovations where participation is imbalanced and dialog is not transparent. Therefore,
two new institutional innovations might help such as: l) incentive-driven performance
contracts between the government and cotton companies with transparent reporting of
results, and 2) the transfer of government shares in JV Cs to farmers, with support from
cotton trust to act on behalf of the new farmer shareholders.

Elimination of price fixing by government would remove one source of intense
conflict in the sector, and would also allow IAM to spend more time on productive

activities to improve sector performance.

2. Establish a favorable GMO policy. Currently, GM seeds are forbidden by law in
Mozambique. The adoption of Bt cotton depends on the legal approval of
commercialization of GM seed. Mozambique does not have a comprehensive biosafety
regulatory framework in place and is unlikely to have one in the near future (Wally,
2004). The recently created working group for this purpose has only three years of
existence and it still in its infancy. To have a full biosafety regulation in place is a

challenging activity and requires a considerable amount of time and effort so, there is a

-107-

need to develop a basic framework for a pragmatic approach that will allow forward
movement before a full biosafety framework is in place. This approach should consist of:

(a) Initial identification of priorities: This part of the framework will help to
ensure that resources are dedicated to biosafety procedures that best serve the country’s
most relevant need. The step-by-step basis could be adopted, including confined small-
and large field trials of GMOs, and finally unconfined release of a GMO. The step
forward would be to recognize that biotechnology in nonfood cash crop like cotton raises
fewer controversial issues than it does in maize, and to move forward with a framework
that would allow the testing of Bt cotton.

(b) Design of framework for implementing biosafety: In this phase, the recently
formed Biosafety working group should benefit from McLean et al.’s (2002) paper that
lays out the linkage between policy, capacity, and regulation based on the experiences
from other countries in Africa. Regarding regulation, the biosafety working group should
bear in mind that regulation does not need to be a restrictive activity, rather a protective
one, information provider, and development facilitator of a certain area.

(c) Identification of priorities for training and capacity building: The steps (a)
and (b) should illuminate needs for training and capacity building. In this phase, the task
would to be identify specific people in Mozambique who have required the skills, and
find resources. The UNEP-GEF Project on Development of National Biosafety
Frameworks is one potential source of long-term support for development of national

biosafety capacity.

—108-

3. Establishing a “closed-loop” system. Under current practice, farmers are not
explicitly charged for cotton seed. This practice would need to change if Bt cotton is to
be introduced. The “closed-loop” system intends to provide assurance to the GM
company, through an organized system, that the seed would not be replanted by farmers,
instead bought every year. The current input delivery system consists of cotton
companies delivering inputs to farmers. The cotton seed would be distributed in the same
way other inputs are delivered currently, but farmers would need to be educated in such a
way to recognize that the cotton seed is not free. This would ensure that the GMO
company or its partner in the country trace all Bt cottonseed by-product from ginning
through to final use.

If Mozambique could ensure a closed-loop system, and give regulatory approval,
then the country could access Bt seed. But, the challenge is that establishing a “closed-
loop” system requires a fairly high level of institutional sophistication — good
organization and management, good use of information — which Mozambique will not
easily attain. So, just as the country needs to take a step-by—step, pragmatic approach on
biosafety regulation (as indicated above), in this case moving forward just on one crop, it
needs to take a step-by—step approach in creating and operating a closed-loop system. This
approach could be implemented under focused regulations that allow testing of Bt cotton,
then by experimentation in farmer fields along with testing of a closed-loop system.
Hence, this approach would benefit from the available research project designed by
University Eduardo Mondlane (UEM) and cotton company (AGRIMO) with the objective

to test Bt cotton (evaluating its production costs and adaptability).

-lO9-

7.4 LIMITATIONS OF THE STUDY

This study relies on a dataset that was collected with the desire to understand the
effects of cash-cropping on smallholder income, not with the objective of comparing
alternative pest management strategies, which is the focus of this study. Consequently, it
gathered partial information on pest management, and as result, that two functional forms
that seem most suitable from theory generated opposite coefficient signs inconsistent with
expectations, forcing reliance on the simpler Cobb-Douglas model. Data specifically
collected to assess pesticide productivity would improve the empirical results of the
study. The estimated average yield gain that turns Bt cotton attractive to farmers is 6%,
well within the confidence interval around the BOLLW yield function coefficient of
0.150 (95% c.i. = [0.023 — 0283]).

The appropriate dataset for this study should have had information on pest
management practices, including what material was sprayed in each spray (quantity of
active ingredient); what was the target insect, timing of each application, and pest

pressure at the time of application.

In addition to data limitations, it is likely that estimated Bt cotton yields are
affected by the environment in which the estimates were generated. These data were
collected under actual farmer practices, which are known to be far less than optimal with
regard to pest control. Specifically, farmers at the time of this data collection had
relatively little knowledge about scouting and proper timing of pesticide application. As

a result, the marginal productivity of pesticides in this data set is likely to be much lower

~110-

than it would be under optimal practices — and lower than that achieved by Bt cotton

which has continual presence of the toxin.

The final judgment of the impact of Bt cotton has to take into account other factors
not included in the analysis such as the biosafety risks that Bt cotton may have, including
the non-market effects on health and environment, including the impact on the ecosystem.
Edge et al., (2001) for instance indicate that the reduction of broad-spectrum insecticidal
sprays reduces harm to beneficial insects. These authors also indicate that the reduction in
insecticide use due to Bt cotton reduces the direct contact of birds and other types of
wildlife with insecticides, runoff of insecticides, air pollution, and related waste.
Therefore, more focused research is still needed to firlly assess the economic,

environmental, and social benefits and risks of Bt cotton as discussed in the next section.

7.5 RECOMMENDATIONS FOR FURTHER RESEARCH

The analysis in this paper implicitly assumes that the Bt traits are integrated into
existing cotton varieties in Mozambique. In fact, integration is not likely to happen
unless the country pays a fee to a company to do this work. Zilberman (2004) notes that
when the local variety is modified genetically, the combination of the genetic
modification and chemicals will reduce pest damage, thus increasing output. Therefore,
there is a need to evaluate what the likely value of such a fee would be and whether it
would still be profitable for the country to pursue Bt.

These existing Mozambican cotton varieties are hairy and thus have reasonably

-111-

good resistance to sucking insects, but their yield potential is very low. Therefore, it is
worthwhile to determine the profitability of Bt cotton if better varieties were available for

integrating the Bt trait.

More research into why plantings are so often delayed in farmer fields would also
have a high payoff. As we estimated, the marginal product for a one week delay in
planting is equal to 94 Kg of yield loss, which corresponds to a financial loss of about
$10/week at a farm gate price of $0.11/Kg. Farmers planting two or three weeks late have
yield losses of 39% and 52%, respectively. Given that the average farmer in Mozambique
performs crop management practices such as land preparation, planting, weeding and
spraying with 30 days of delay, the impact of late seeding becomes an important factor to
consider. Many factors could be determining the seeding date, and so more research is
needed. This research objective would fit well with the new socio-economics unit to be
established in the consolidated agricultural research institute, Instituto de Investigacao
Agrdria de Mogambique (11AM). .

Producing cotton in Mozambique is almost impossible without the use of
pesticides, which poses serious environmental impacts. It would be useful to complement
this study by evaluating the environmental impact of pesticides, given that pesticide
savings from Bt cotton use could result in an environmental benefit for the society.
Environmental concerns on Bt cotton are gene flow to wild relatives that can possibly
result in economic consequences since crop-to-weed flow can promote the evolution of
more aggressive weeds. So, we recommend assessing the possible harmful effects on the

environment.

-112-

Although the pesticide cost in Mozambique to farmers is lower than its economic
cost, farmers still use these chemicals at less than optimal rates. Some possible reasons
are discussed in this thesis but none of them are scientifically tested in the case of
Mozambique. So, we recommend that research be done on factors determining the low

insecticide application rates.

This study did not perform a risk analysis of the probability of various outcomes
from the adoption of Bt. We recommend that to complete this research, a risk analysis

could be carried out.

Finally, according to Schmitz (2004), when assessing benefits and costs of GM
crops as we do, the size of segregation costs and trade implications are important. Our
results indicate that Bt cotton is attractive from society’s point of view. However,
Mozambique exports lint to European Union (EU) members, which have strict treatment
of GM crops, and so logical questions are: 1) to what extent will the adoption of Bt cotton
in Mozambique result in possible trade “barriers” from EU, and 2) what is the magnitude
and nature of gains expected from trade. This is one potential area for further research
because its outcome will help to decide whether or not Mozambique will gain from trade
by adopting GM technology. Tothova and Oehmke (2004) observe that for food-insecure
countries, the trade implication can force tradeoff between maintaining export market and

food availability.

-113-

APPENDIX A

World Price Projections

-114-

Figure A-l Cotton Price Projections

 

 

 

 

 

 

 

1600-
1500“ "a.
---|ndexA .'
? —U.S.Farm ,v’
ol400+
'—
.9
gwom
«I»;
d)
@1200-
a.
1100*
1000 I I I I I I I I I I
o N '1, '5 u 6: a: '\ a: 9 c
Q S \\ a) n} r} 6 o «\ a) \'\
q\ a o a o c a o o o <2.
.39 «s «s '19 as e «e «s e «9.8%

Cropping Season

Considering the average cotton price projected in the next 6 years ($1,423.80/ton),

result in farmer price in Mozambique of $0.63/Kg.

-115-

Figure A-2 Cotton Price Trend between 1990 and 2000.

 

 

 

 

500 -
- - - lndexA

450* . ($IMT)

400 J .' . —MOZ($/MT)

350 «

300 «

N

o

o
r

 

Price ($lMetric Ton)
N
0|
0

 

 

I

1988 1990 1992 1994 1996 1998 2000 2002
Years

Formulae used to calculate the farm gate cotton price in Mozambique

Cotton price (Mt/Kg) = {(INDEX A — Qdiff.)*COVER *OER + SEEDCOST}*55%
Where:

Qdiff. = quality difference based on cropping (for season 1999/00 = 4.12 cents/Lb)
OER = official exchange rate,

SEEDCOST = 1.2 Kg of seed (by product) evaluated at 850 Mt,

CONVER = conversion factor from Lb to Kgs (2.22046),

Freight and insurance is assumed to be 10 cents/Lb

-116-

APPENDIX B

Questionnaire

-117-

Ministério de Agricultura e Desenvolvimento Rural
e
Universidade Estadual de Michigan

Percepgoes sabre Algodao Transgénico (“Br cotton”)

Entrevista aos informantes-chave

[Melhoradores, empresas de Agroquimicos, empresas algodoeiras e
pesquisadores]

J ulho - Agosto de 2003

AVISO

A sua participacao nesta intrevista é inteiramente voluntaria. Qualquer informacao que
fomecer sera tratada com sigilo e confidencialidade de acordo com o Decreto 12/82 de 22 de
Julho de 1982 e, em nenhuma circunstancia o seu nome sera associado a nenhuma resposta.

 

VARIAVEIS DE IDENTIFICACAO

Provincia PROV

Distrito DIST

Nome da empresa COMPAN

Tipo de empresa SERV l-pesquisa/entomologia, 2-Agroquimicos,

3-empresa algodoeira, 99-outra (especificar)
(Marque com circulo)

Nome do respondent RNAM S/Funcao:

Data da entrevista SURDAT

As Plantas geneticamente modificadas constituem um centro de acesos debates da actualidade
acerca da utilidade da agricultura biotecnologica (Pray and Ma, 2001). Enquanto as grandes
empresas em biotecnologia como a Monsanto e as organizacoes de industrias biotecnologicas
acreditam que a agricultura biotecnologica pode ser uma altemativa para resolver os problemas de
pobreza bem como 05 de uso intensivo de pesticidas, os contra biotecnologia argumentam que

plantas geneticamente modificadas constituem um perigo para o meio ambiente e n50 estao
desenhadas para ajudar os pobres se tiver em conta o seu preco. 0 abiectivo do presente estuda e

de calher a sua agim'aa sabre alguns assuntas gue galarrzam estes debates bem coma ('udar a
nerceber melhar a passive! imgacta gue a algodao transgenico gadera ter em Mocambigue.

 

-118-

QAAI. Por favor, poderia descrever sumariamente a sua empresa no que concerne a:
a. Ano de fundacao

b.Que servicos providencia aos camponeses?
c. Alguma vez a sua empresa teve que ajustar a sua estrutura para melhor responder as
necessidades do campones? Que ajustamentos fez?

 

 

 

QEAA6. De acordo com a sua experiencia, quais $50 as pestes mais probleméticas no algodao
em Mocambique e como poderia minimizar o seu efeito?

(fibendo-se que o efeito das pragas pode variar consoante as condicoes aggo-climéticas, p_o;leria
fomecer informacao para cada tipo de estacgio: Os tipos de estagao sao: Ana normal, seco e

ht'rmido)
Tabela

 

 

oara Ano normal

Periodo de
incedéncia

Danos
causados
(max.
rendimento
perdido)

Condicoes
favoraveis

Locais de
alta
incidéncia
(distritos)

Pesticidas
sao
apripria-
dos

  

Quando
aplicar
pesticidas
?

 

Tabela

 

  

 

oara Ano Seco

 

Periodo de
incedéncia

 

Danos
causados
(max.
rendimento
perdido)

 

Condi-cées
favoraveis

 

Locais de
alta
incidéncia
(distritos)

 

 

Pesticidas
apropriad
os

 

Quando
aplicar
pesticidas
?

(Semana e
més)

 

 

 

 

-Il9-

 

 

 

 

Tabela para ano chuvoso

Periodo de Danos Condi-coes Locais de Pesticidas Quando
incedéncia causados favora-veis alta apropriad aplicar
(max. incidéncia os pesticidas
rendimento (distritos) ?
perdido)

 

 

 

 

 

 

 

 

QAA8.Quais 550 05 pesticidas que a sua empresa importa e/ou distribue aos camponeses?

 

Nome do pesticida Classe (I, II, III or IV) Conteudo de ingrediente activo I

 

 

 

 

 

2M9. Quais os custos envolvidos na importacao de agro—quimicos?

 

Custos Preco par ton

 

Preco FOB em (U.S. S)

 

Frete para 0 local de importacao (US. 3)

 

Desembarque no local de importacao (US. $)

 

Seguros

Custos de porto local

 

 

Custos de transporte e comercializacao ate 0 mercado grossista

 

Custos de transporte e comercializacao até o mercado retalhista

 

Armazenamento

 

 

 

 

FALEMOS AGORA SOBRE O ALGODAO TRANSGENICO
QEAABa. Tera ouvido falar de algodao transgénico?
QEAA3. Quais 550 as expectativas em relacao ao desempenho do algodao transgénico ("Bt

cotton ") em
Mocambique?

 

-120—

QE4. Tern alguma ideia de variedades que possam estar adaptadas a condicoes de Mocambique
ou similares? (Por favor fomeca informacao mais detalhada sobre elas na tabela abaixo)

Preco da
semente

Nivel de
resisténcia a

Exigén-
cias

Pesticidas
em mais

pestes.

pesteci-das

apropria-

desvanta
-gens

0 nenhum (no. de dos

aplica-
coes)

l baixo,
2 médio
3 alto

 

 

 

 

 

 

 

 

 

QES. O que poderia recomendar para minimizar o efeito de resisténcia das pestes ao algodao
transgénico (“BI cotton ")?

l. Cultura hospedeira

2. Alta dosagem do CryAl na formulacao das variedades

3. Ambas

4. Outra (especifique)

 

Justifique as suas opcoes:

 

QEAAGole. Em caso de se passar a produzir algodao transgénico em Mocambique, quais
seriam as vantagens, desvantagens que dai poderiam surgir, como é que a sua empresa seria
afectada e como poderia reagir face ao choque?

Govemo
(Instituto de
Algodao)

Pesquisa-
dores

Empresas de
Agro-quimicos

Empresas
algodoeiras

 

 

 

Solucoes

 

 

 

 

 

 

-121-

QAAll. Quais seriam as condicoes necessérias para a sua empresa adoptar o algodao
transgénico?
QAA12. Se 0 camponés tivesse que pagar o custo real de cada servico prestado pela sua
empresa, quanto lhe custaria:

a. Pulverizadores

b. Pesticidas

c. Servicos de extensao

d. Cre'dito

e. Outros

 

 

 

 

QAA13. Qual destes servicos a sua empresa subsidia ao camponés e qual e o valor subsidiado?
a. Pulverizadores
b. Pesticidas
c. Servicos de extensao
d. Crédito
e. Outros

 

 

 

Q14. Comentarios:

 

-122-

APPENDIX C

Selection bias test

-123-

Table C-l Cotton Yield Regression coefficients in Northern Mozambique, 2003.

—
Variable Equation 130 Equation 23]

Coef. S. E. gvalue Coef. S. E. p-value

 

1n (TOTWAE) 0.44 1.289 0.001 0.43 0.129 0.001
In (HERB) 0.62 0.269 0.024 0.56 0.263 0.036
In (MAIZE YIELD) -0.08 0.082 0.333
VILMTZl 1.46 0.590 0.015 1.45 0.590 0.015
VILMTZZ 1.53 0.612 0.014 1.52 0.612 0.014
VILMTZ4 1.07 0.612 0.084 1.07 0.612 0.082
VILMTZS 2.02 0.704 0.005 2.07 0.702 0.004
VILMTZ7 1.23 0.624 0.051 1.28 0.621 0.042
VILMECZ 1.74 0.574 0.003 1.74 0.574 0.003
VILMONI 1.71 0.625 0.007 1.72 0.624 0.007
VILMON4 1.89 0.601 0.002 1.89 0.602 0.002
VILMONS 1.54 0.680 0.025 1.57 0.680 0.023
VILMON6 2.05 0.824 0.015 2.06 0.824 0.014
Constant 4.03 1.18 0.001 3.61 1.09 0.001
Ln (YIELD)= Dep. Var. N=122, F-value =5.79, N=122, F-value = 6.02,
p-value=0.000 p-value = 0.000
Adj. R-square = 0.4767 Adj. R-square =0.4769

 

Source: Author’s computations based on Strasberg (1997)

 

From the results presented in the two regressions we observe that maize yield
(instrument for selection bias) is not significant although has negative sign. Thus, we

decide not to worry much about selection bias.

 

3° Variance-Weighted Least Square Regression of Cotton yield excluding maize yield.
3' Variance-Weighted Least Square Regression of Cotton yield including maize yield.

-124-

APPENDIX D

Village Identification

-125-

Table D-l

Definition of the village dummy variables

 

 

 

— —
Variable District Village Name Frequencel’z
VILMTZl Montepuez Macarranga 30
VILMTZZ Montepuez Nacuaia 12
VILMTZ3 Montepuez Nacuca 37
VILMTZ4 Montepuez Nropa/Mondiane l 5
VILMTZS Montepuez Nacimoja 5
VILMTZ6 Montepuez 25 de Setembro 4
VILMTZ7 Montepuez Linde 10
VILMECl Meconta Napipine 17
VILMECZ Meconta Varrua 1
VILMEC3 Meconta Napita 2
VILMONl Monapo Mepine 2 l
VILMON2 Monapo Natete l 3
VILMON3 Monapo 3 de Fevereiro 20
VILMON4 Monapo Namacopa l 8
VILMON5 Monapo Nacololo 7 '
VILMON6 Monapo Picadane 3

 

Source: Adapted from Strasberg (1997).

 

’2 Total of 215 cases.

-126-

 

 

 

 

Table D-2 Regression coefficients for village dummy variables in Northern
Mozambique, 2003.

Quadratic Translog Cobb-Douglas

Variable Coef. S. E p- Coef. S. E p- Caef. S. E p-
value value value
Vilrntzl -168.7 221.7 0.448 0.19 0.360 0.598 0.41 0.372 0.273
VilmtzZ -128.0 240.0 0.595 0.21 0.39 0.588 0.62 0.407 0.132
Vilmtz4 -370.1 239.5 0.201 -0.06 0.386 0.886 0.12 0.400 0.757
VilmtzS 93.4 259.3 0.719 0.79 0.425 0.065 1.05 0.441 0.019
Vilmtz6 -543.7 282.3 0.056 -0.98 0.459 0.034 -0.97 0.479 0.045
Vilrntz7 -1 60.6 243.9 0.511 0.17 0.407 0.676 0.50 0.421 0.239
Vilrnonl -157.7 225.1 0.484 0.24 0.363 0.499 0.43 0.378 0.253
Vilmon2 -354.1 458.4 0.441 -0.46 0.722 0.524 -0.53 0.754 0.483
Vilmon3 -371.9 221.7 0.095 -0.32 0.355 0.373 -0. 18 0.372 0.634
Vilmon4 28.2 226.5 0.901 0.29 0.363 0.428 0.65 0.372 0.081
VilmonS -1 10.2 243.4 0.651 0.36 0.393 0.364 0.59 0.409 0.154
Vilmon6 -179.4 279.3 0.522 0.40 0.452 0.378 0.57 0.472 0.227
Vilmec2 35.7 221.6 0.872 0.53 0.354 0.136 0.68 0.370 0.069
Vilmec3 -301.4 233.8 0.199 -0.24 0.372 0.525 0.01 0.385 0.970
Dependent Adj R2 = 0.5561 F(31,183) Adj R2 = 0.5408 Adj R2 = 0.4925
Variable is =12.66 Prob>F= 0.0000 F(30,184) = 11.50 F(24,l90) = 9.60

Yieldian/ha Prob>F = 0.0000 Prob>F = 0.0000

 

Source: Author’s computations based on Strasberg (1997).

 

-127-

APPENDIX E

Mackinnon, White and Davidson test (MWD) for Non-nested models

-128-

To illustrate the test, assume the following:

Ho: quadratic model: Yield is a function of regressors, the X’s

H1 Translog model: lnYield is a fimction of logs of regressors, the logs of X’s

Step 1: We estimated the quadratic model and obtained predicted values, ho.

Step 2: We estimated the translog model and obtained its predicted values, lnhl

Step 3: we computed z] = My lnho

Step 4: regressed of log Yield on X’s and Z] in the translog model. We rejected Ho if the
coefficient 21 is statistically different fiom zero by the usual t-test, then the translog is not
the true model.

Step 5: we computed z; = (antilog of lnhl-ho)

Step 6: regressed the Yield on X’s and 22 on quadratic model. We rejected H1 if the
coefficient 22 is statistically different from zero by the usual t-test, then the quadratic is
not the true model. The results are summarized in the table F -1.

In both regressions 21 and 22 are statistically significant. Therefore, no one model
is the true one. Both cannot be rejected. Then, to select the superior the Adj. R2 was be
used. But because they have different measurement units, the anti log of the predicted
yield from translog was correlated to the observed yield and, Woodridge (2000) suggests
to calculate the comparable R2 of translog by taking the square of the correlation
coefficient between h1 and yield, resulting in 0.6296 (0.79352). Then, the R2 for translog

is greater than that of the quadratic model of 0.5777. So, we pick the translog model.

-129-

Table E-l Quadratic and Translog regressions for MWD test

 

 

Variable Quadratic Translog
Coef. S. E. p—value Coef. S. E. p—value

TOTWAE -1.96 0.949 0.041
TOTWAE2 0.32 0.128 0.012
SEEDWEEK 3.46 1.079 0.002
SEEDWEEK2 -15.51 7.612 0.043 -1.14 0.317 0.000
FERTQ 23.64 12.72 0.065

INSPESTS -356.1 1 91.131 0.000 -0.27 0.142 0.062
LANDQUAL -82.89 3 8.106 0.031 -0.022 0.098 0.025
BOLLW -1.24 0.278 0.000
BOLLW 2 26.84 10.165 0.009 0.87 0.170 0.000
FERTWSEED -1.48 0.879 0.095

HERBQ 0.22 0.060 0.000
VILMTZS 1.19 0.431 0.007
VILMTZ6 -557.37 306.95 0.071

VILMON5 0.75 0.400 0.061
VILMECZ 0.86 0.360 0.018
Zr 0.01 0.000 0.004
22 214.67 109.39 0.051

Constant 5.13 1.88 0.007

 

N=215, F-value = 10.36,
p-value = 0.000
Adj. R2 = 0.5843

N=215, F-value = 10.10,
p-value = 0.000
Adj. R2 = 0.5686

 

Source: Author’s computation based on Strasberg (1997).

-130-

APPENDIX F

Pesticide Endogeneity - Hausman test

-131-

Table F-l Regression coefficients for BOLLW endogeneity test

—
Variable Regression 11’3 Regression 234

Coef. S. E. p-value Coef. S. E. p-value

 

In (TOTWAE) 0.42 0.098 0.000
In (SEEDWEEK) -0.36 0.163 0.030
In (FERTQ) 0.10 0.045 0.029
INSPESTS 0.42 0.165 0.01 1
In (LANDQUAL) -0.41 0.210 0.051
In (BOLLW) 0.14 0.065 0.033
Residual 1.54 1.53 0.315
In (SUCKINS) -0.22 0.09 0.025
VILMTZS 1.44 0.592 0.016
VILMTZ6 -1.38 0.62 0.028
VILMONI 0.77 0.420 0.067
Constant 2.81 1.34 0.038

N=215, F-value = 1.99, N=215, F-value = 9.30,

p- value = 0.0060, p-value = 0.000,
Adj. R-square = 0.0995 Adj. R-square = 0.4923

Dep. Var. = 1n (BOLLW) D52. Var. = In (YIELD)
Source: Author’s cor-rmutations based on Strasberg (1997)

The coefficient of the residual is not significant. Therefore, we conclude that insecticide

applications to control bollworms are not endogenously decided by the farmers.

 

33 Regression with BOLLW as dependent variable as function of all other exogenous variables besides the
usual explanatory variables.
3" The original regression including the predicted BOLLW as an additional explanatory variable.

~132-

APPENDIX G

Yield Estimations

-133-

According to Verbeek (2000), predicted yield from logY = f(x) is underestimated if
set that is equal to exp (log xi’) because taking log is a nonlinear transformation. The only
way to get around this problem is to make distributional assumptions. Assuming for
example that the error term is normally distributed with mean zero and variance 52; it will
imply that the conditional distribution of yi is lognormal. Therefore, to adjust for the
nonlinear transformation, the half-variance term is added, assuming that the error term is
normally distributed. So, the yields were estimated as described below.

To illustrate how yield values with and without Bt were estimated, consider for
instance, seeding week three (SEEDWEEK=3) with no additional insecticide application
under low-input use (assumptions: no fertilizer and herbicide use), higher pest infestation
(INSPESTS=1), and fertile soils (LANDQUAL=1).

We start by presenting the final Cobb-Douglas model that will be used to predict
yields with Bt cotton as:

(1) Conv. Cotton Yield = exp"{4.76 + Zvillage*freq35/n + 0.43*ln(TOTWAE) -

0.38*1n(SEEDWEEK) - 0.29“'INSPESTS + 0.11*ln(FERTQ) + 0.15*ln(BOLLW) +

0.12*ln(HERBQ) - 0.22*ln(LANDQUAL) + 0.121n(AGE) + 0.04*1n(SUCKI) + s2*0.5}
where s is sample standard deviation (0.270391) and s2 is sample variance.

Using this model, the conventional cotton yield under the scenario stated above
can be calculated by inserting representative values for major inputs. As we mentioned in
section 3.3, the representative values represent the most common values and sample
average values in the sample for continuous and integer variables, respectively. For

instance, the sample average for TOTWAE is 49 days, for AGE is 35 years, for F ERTQ is

21Kg/ha, for HERBQ is 8.3 litres/ha, while the most common values for BOLLW is 3

-134-

insecticide applications (35% of the sample applied insecticide three times to control
bollworms). To assure that three insecticide applications are in fact a representative
figure, we appeal to MADER (2000a) in the same study area where we got the same
figure. For SUCKI, the representative value is 1 (29% of the sample), and for
LANDQUAL is 1(61% of the sample).
(Eq G1) Conv. Cotton Yield = exp {4.76 + 0.36 + 0.43*ln(49) - 0.38*1n(3) -
0.29*1 + 0.11*ln(0+1) + 0.15*1n(3) + 0.12*ln(0+1) - 0.22*ln(1) + 0.12*1n(35) +
0.04*ln(0+1) + 02703912405}
= exp"(6.725) = 833 Kg/ha

To calculate the conventional cotton yield, we inserted BOLLW=3 in the Eq Gl
because under typical practice, farmers apply mostly three times to control bollworms.

Predicted Bt cotton yield was estimated the same way, but BOLLW=6, to simulate
cotton yield with six insecticide applications to control bollworms for complete bollworm
control, which is the principal characteristic of Bt cotton:
(Eq G2) Conv. Cotton Yield = exp" {4.76 + 0.36+ 0.43*ln(49) - 0.38*ln(3) -
0.29*1 + 0.11*1n(0+1) + 0.15*ln(6) + 0.12*ln(0+1) - 0.22*ln(1) + 0.12*35 +
02703912405}
= 926 Kg/ha

Then, we reduced the Bt cotton yield by an amount equivalent to the effect of one
insecticide application to control sucking insects. This adjustment accounts for the fact
that the current Bt cotton varieties appear to be less resistant to sucking-insects than the
local conventional cotton varieties. The cotton yield effect of one insecticide application

for sucking-insect control was estimated as the difference between conventional cotton

yields with and without one insecticide application:

 

3’ See Appendix D.

-l35-

(Eq G3) SUCKI Yield effect = Conv. Cotton Yield (Eq G1) - Conv. Cotton Yield with
one SUCKI

Where,

Conv. Cotton Yield with one SUCKI = exp"{4.76 + Zvillage*fi'eq/n +
0.43*1n(TOTWAE) - 0.38*1n(SEEDWEEK) - 0.29*INSPESTS + 0.11*1n(FERTQ) +
0.15*ln(BOLLW) + 0.12*1n(HERBQ) - 0.22*ln(LANDQUAL) + 0.121n(AGE) + 32*05 +
0.04*ln(_2_)}

One could ask why 0.004*ln(2) instead of 0.004*1n(1). This is simply because the
log(1)=0 and this results that Yield of conventional cotton in Eq 61 becomes equal to
conventional yield with one SUCKI, resulting that the SUCKI Yield effect in Eq G3 is
equal to zero (since ln(l) = 0, then the effect of SUCKI in the yield estimation
disappears). So, to handle this methodological problem, ln(2), was used to reflect the idea

that if conventional cotton still use SUCKI=1, the effect of one addition SUCKI will

result in some positive yield difference over the conventional cotton yield with SUCKI=1.

Conv. Cotton Yield with one SUCKI= exp"{4.76 + 036+ 0.43*ln(49) - 0.42*1n(3) -
0.29*1 + 0.11*1n(0+1) + 0.15*In(3) + 0.12*ln(0+1) - 0.22*ln(1) + 0.12*35 +
0.2703912*0.5 + 0.04*ln(2)} = 858 Kg/ha. Using the equation (3) the SUCKI Yield effect
is given by 858 — 833 = 25Kg/ha.

Based on the -25 Kg/ha predicted yield effect of current Bt cotton varieties’
greater susceptibility to sucking insect damage, the likely Bt cotton is calculated as:
(Eq G4) Bt Cotton Yield = (Eq GZ)-(Eq G3) = 926 — 25 = 901 Kg/ha.

The percentage Bt cotton yield gain over conventional cotton is calculated as:
(Eq G5) Bt cotton yield gain = 100*[(Eq G4)-(Eq G1)]/(Eq G1), or simply 100*(901-

833)/833 = 8.2%.

-136-

For other scenarios, the yields were estimated the same way, and table 4-5 reports

values obtained in the steps described in Equations (G1), (G4), and (G5).

-l37-

APPENDIX H

Financial Partial Budget

-138-

Table H-l

 

Financial Bt cotton Partial Budget

Bt Cotton Returns by seeding week

 

 

 

 

 

Level of input use Items (Bt - Conventional cotton)
3 4 5 6 7 8
seed cotton yield (Kg/ha) 23 21 19 18 16 15
seed cotton price($/ha) 0.1 1 0.1 1 0.1 1 0.1 1 0.1 1 0.1 1
Gross revenue (8013) 2.52 2.28 2.10 1.93 1.72 1.67
N0, , insecticide -994 -994 -994 -994 -994 -994
additional seed cotton 16.60 16.60 16.60 16.60 16.60 16.60
”5?“: Spraying labor ($lha) .405 .405 .405 .405 .405 .405
app I“ on family ae labor days 57.00 57.00 57.00 57.00 57.00 57.00
Net returns to land ($lha) -0.09 -0.33 -0.52 -0.68 -0.89 -0.94
Net returns to fam. labor
(S/ae) 0.00 -001 -001 -001 -002 -002
seed cotton yield (Kg/ha) 6 4 4 5 4 4
seed cotton price($/ha) 0.1 1 0.1 1 0.1 1 0.1 1 0.1 1 0.1 1
Gross revenue (S/ha) 0.65 0.42 0.46 0.52 0.42 0.38
insecticide -10.10 -10.10 -10.10 ~10.IO -lO.lO ~10.10
One . seed cotton 16.60 16.60 16.60 16.60 16.60 16.60
additional .
pesticide Spraying labor (S/ha) -4.05 -4.05 -4.05 -4.05 -4.05 -4.05
application farruly ae labor days 57.00 57.00 57.00 57.00 57.00 57.00
Net returns to land (S/ha) -l .80 -2.03 -1.98 -1.92 -2.03 -2.06
Net returns to fam labor
($/ae) -0.03 -0.04 -0.03 -0.03 -0.04 -0.04
seed cotton yield (Kg/ha) -8 -7 -6 -6 -6 -5
seed cotton price($/ha) 0.1 1 0.1 1 0.1 1 0.1 1 0.1 1 0.1 1
LOW " Gross revenue (S/ha) -0.88 -0.78 -0.62 -0.65 -0.64 -0.57
“PUT TW‘? . insecticide -1027 -1027 -1027 -1027 .1027 -1027
USE add‘i‘i’m seed cotton 16.60 16.60 16.60 16.60 16.60 16.60
pestrcrde .
applications family ae labor days 57.00 57.00 57.00 57.00 57.00 57.00
Spraying labor (S/ha) -4.05 -4.05 -4.05 -4.05 -4.05 -4.05
Net returns to land (Slim) -3. l 6 -3.06 -2.90 -2.93 -2.92 -2.85
Net returns to fam. labor
(S/ae) -0.06 -005 -005 -005 -005 -005
seed cotton yield (Kg/ha) -17 -15 -l4 -13 -12 -11
seed cotton price($/ha) 0.1 1 0.1 l 0.1 1 0.1 1 0.1 l 0.11
Gross revenue ($/ha) -1.85 -1.63 -1.57 -1.39 -1.37 -1.18
““9? insecticide -1044 -1044 -1044 40.44 -1044 -1044
add‘F‘f’m‘ seed cotton 16.60 16.60 16.60 16.60 16.60 16.60
”Wad? family ae labor days 57.00 57.00 57.00 57.00 57.00 57.00
applications .
Spraying labor (SM) 405 .405 .405 .405 .405 .405
Net returns to land ($/ha) -3.97 -3.74 -3.68 -3.50 -3.48 -3.30
Net returns to fam. labor
(S/ae) -0.07 -0.07 -0.06 -0.06 -0.06 -0.06

 

-139-

 

 

 

 

Table H-l (Cont’d).

seed cotton yield (Kg/ha) 42 38 34 32 30 29
seed cotton price($/ha) 0.1 l 0.1 I 0.1 l 0.1 l 0.11 0.11
No Gross revenue (S/ha) 4.66 4.20 3.78 3.53 3.28 3.17
additional Insecticide -9.94 ~9.94 -9.94 -9.94 -9.94 -9.94
pesticide seed cotton 16.60 16.60 16.60 16.60 16.60 16.60
application family ae labor days 57.00 57.00 57.00 57.00 57.00 57.00
Spraying labor (Slha) -4.05 -4.05 -4.05 -4.05 -4.05 -4.05
Net returns to land ($/ha) 2.05 1.59 1.17 0.92 0.67 0.56

Net returns to fam labor
($/ae) 0.04 0.03 0.02 0.02 0.01 0.01
seed cotton yield (Kg/ha) 9 9 8 8 6 7
seed cotton price($/ha) 0.1 1 0.1 1 0.1 1 0.1 1 0.11 0.11
One Gross revenue ($lha) 1.04 0.94 0.86 0.84 0.70 0.72
additional Insecticide -10.10 -10.10 -10.10 -10.10 -10.10 -10.10
pesticide seed cotton 16.60 16.60 16.60 16.60 16.60 16.60
application Spraying labor ($/ha) -4.05 -4.05 -4.05 -4.05 -4.05 -4.05
family ae labor days 57.00 57.00 57.00 57.00 57.00 57.00
Net returns to land ($lha) -1.41 -1.51 -1.59 -1.61 -1.74 -1.73

HIGH- Net returns to fam labor
INPUT jS/ae) -002 -003 -003 -003 -003 -003
USE seed cotton yield (Kg/ha) -14 -12 -12 -1 1 -10 -9
seed cotton price($/ha) 0.1 1 0.11 0.1 1 0.1 1 0.11 0.1 1
Two Gross revenue (5013) -1.57 -1.33 -1.28 -1.18 -1.08 -0.96
additional Insecticide -10.27 -10.27 -10.27 -10.27 -10.27 -10.27
pesticide seed cotton 16.60 16.60 16.60 16.60 16.60 16.60
applications family ae labor days 57.00 57.00 57.00 57.00 57.00 57.00
Spraying labor (S/ha) -4.05 -4.05 -4.05 -4.05 -4.05 -4.05
Net returns to land (S/ha) -3.85 -3.61 -3.56 -3.46 -3.36 -3.24

Net returns to fam labor
(S/ae) -007 -0.06 -0.06 -0.06 -0.06 -0.06
seed cotton yield (Kg/ha) -31 -32 -25 -24 -22 -22
seed cotton price($/ha) 0.1 1 0.1 1 0.1 1 0.1 1 0.1 1 0.1 1
Three Gross revenue (S/ha) -3.39 -3.55 -2.74 -2.63 -2.42 -2.39
additional Insecticide -10.44 -10.44 -10.44 -10.44 -10.44 -10.44
pesticide seed cotton 16.60 16.60 16.60 16.60 16.60 16.60
applications family ae labor days 57.00 57.00 57.00 57.00 57.00 57.00
Spraying labor ($/ha) -4.05 -4.05 -4.05 -4.05 -4.05 -4.05
Net returns to land (Slha) -5.51 -5.66 -4.85 -4.74 -4.53 -4.51

Net returns to fam labor
(S/ae) -010 -010 -009 -0.08 -0.08 -0.08

 

Source: Author’s comutations based on Strasberg (1997), MADER (2000), and MADER (2001a).

-140-

APPENDIX I

Economic Partial Budget

-141-

 

 

Table I-l Foreign exchange premium calculation, Northern Mozambique, 2003.
Year 1998 1999 2000 2001 2002
Exports of goods and
services (% of GDP) 10.49 10.12 12.87 21.75 23.51
GDP (current U85) 3.87E+09 3.98E+09 3.68E+09 3.44E+09 3.60E+09
Imports of goods and
services (% of GDP) 27.89 37.77 38.9 35.45 38.23
Exports (US $) 4.06E+08 4.03E+08 4.74E+08 7.48E+08 8.46E+08
Imports (US $) 1.08E+09 1.50E+09 1.43E+09 1.22E+09 1.38E+09
OER 1 1853.42 12673.25 151 13.08 20679.59 23746.51
CPI 100.28 103.42 1 16.58 127.13 1.48E+02
EXPORTS REAL (MT) 3.42E+04 3.07E+04 2.69E+04 2.85E+04 2.40E+04
IMPORTS REAL (MT) 9.08E+04 1.15E+05 8.12E+04 4.64E+04 3.90E+04
Exports duty rate (%)36 0.33 0.33 0.33 0.33 0.33
Imports duty rate (%)37 0.4 0.4 0.4 0.4 0.4
Foreign premium 1.20 1.25 1.22 1.12 1.12
Average 1.18

 

Source: Author’s computations based on IFS (2004), WBI, (2003), NTIS, (2003), and

MADER (2003a).

 

36 Average duty rate for exploitation and exportation of natural resource, cashew nut, fish, and prawns.
37 Duties on imported goods range from zero to 30%. A duty of 30% is levied on consumer goods; and a
value-added tax of 17% is assessed at point of sale. Mozambique does not use import quotas.

-142-

 

 

 

 

 

Table I-2 Economic Seed Cotton Export Parity Price, Northern Mozambique, 2003.
Economic

in U.S. S in Meticals ‘000
“ems Data Lint Seed Lint Seed
CIF (S/ton) 0 0
Freight, insurance, handling (S) 0 0
FOB Port Nacala or Maputo (S/ton) 1059.7 75.8
FOB in '000 Mt/ton converted at 23,963 Mt/U.S.$ 23,963 25,394 1,817
Border Price, Port Nacala or Maputo 1059.7 75.8 25,394 1,817
Adjusted for (1+ foreign exchange premium) 1.18 1250.4 89.5 29,964 2,144
Minus export duty [not deducted in economic analysis]
Minus port handling in Nacala a/
Minus transport from ginner to Port of Nacala a/
= Export Parity price, gin/project site 1250.4 89.5 29,964 2,144
Convert to seed cotton: 0.35 0.65 0.35 0.65
= Seed cotton equivalent 437.7 58.2 10,488 1,394
Combined lint + seed price 495.8 11,881
Company cost factor” 55% 55% 55%
= Export Parity price of seed cotton, farm gate, per ton 272.7 6534.6
= Export Parity price of seed cotton, farm gate, per kg 0.2727 6.535
Implied OER 23,963
Implied costs for port handling, transport, ginningbalingstorage, collection/internal transport: 5,530

 

 

Source: Author’s computations based on MADER (2001b), 1AM (2003).

 

38 Reliable data on costs forced us to use a rule of thumb that cotton companies in Mozambique, in that to
cover costs and earn a "reasonable" profit; they can pay farmers not more than 55% of the FOB price. This
rule is used to propose farm price to IAM in the process of cotton price definition (See Appendix A).

-l43-

Table 1-3 Financial Seed Cotton Export Parity Price, Northern Mozambique, 2003

 

 

 

 

 

Financial

in U.S. S in Meticals ‘000
“ms ”m Lint Seed Lint Seed
CIF (S/ton) 0 0
Freight, insurance, handling (3) 0 0
FOB Port Nacala or Maputo (Slton) 1059.7 75.8
FOB in '000 Mt/ton converted at 23,963 Mt/U.S.$ 23,963 25,394 1,817
Border Price, Port Nacala or Maputo 1059.7 75.8 25,394 1,817
Minus export duty 2.5% 26.5 1.9 634.8 45.4
Minus port handling in Nacala
Minus transport from ginner to Port of Nacala
= Export Parity price, Gin/project site 1033.2 73.9 24758.8 1771.5
Convert to seed cotton: 0.35 0.65 0.35 0.65
= Seed cotton equivalent 361.6 48.1 8,666 1,151
Combined lint + seed price 409.7 9,817
Company cost factor 55% 55% 55%
= Export Parity price of seed cotton, farm gate, per ton 225.3 5399.4
= Export Parity price of seed cotton, farm gate, per kg 0.2253 5.399
Implied OER 23,963
Implied costs for port handling, trangpprt, ginningpaling/storage, collection/internal transflt: 4,601

 

 

Source: Author’s computations based on MADER (20013)i 1AM (2003 i'

Comparing the financial export parity price ($0.23/Kg) to the official cotton price
($0.11/Kg), One can see that the farm gate price is almost 50% less than what farmers
would get if they were incurring all the costs from farm to port. In this situation, farmers

would be better off.

-144-

Table I-4 Economic Bt Cotton Seed import parity price, Northern Mozambique,

2003.

 

Economic cost

 

 

Data in U.S. S in Meticais
Items ‘000
CIF ($/ton) 664.00
Added to Inrportation cost except duties (5) (freight, insurance, 5.2%
handling)39
FOB in ‘000 Mt/ton converted at 23,963 Mt/U.S.$
Adjusted for (1+foreign exchange premium) 1.18
= Border price, port of Nacala per ton 698.31 19,603
Added to the wholesale margin 23%
= Import parity price in Nanrpula per ton 1,006.20 24,112
Added to transportation cost within Nampula districts per ton‘0 139.00 3,331
= Import Parity Price of Bt seed, farm gate, per ton 1.145120 27,443
= Import Parity Price of Bt seed, farm gate per Kg 1-15 27.44
Implied OER 23,863

 

- Source: Author’s conmutations based on MADER/MSU/F SP (2003), Ismail et al. (2001).

 

39 Aggregate cost with importation except import duties. Assumed nontraded.

‘0 Assumed nontraded and lOKm at $13.9/ton/Km.

-145-

Table I-5

Mozambique, 2003.

 

Items

Bt cotton returns by seeding week

(Bt - Conventional cotton)

Partial Budgets to the Economy as whole (Economic Analysis), Northern

 

 

 

 

Level of input use 3 4 5 6 7 3
Cotton yield (Kg/ha) 22.95 20.70 19.05 17.55 15.65 15.20
Gross revenue (S/ha) 6.26 5.64 5.19 4.79 4.27 4.14
Insecticide -28.03 -28.03 -28.03 -28.03 ~28.03 -28.03
No Bt seed 25.00 25.00 25.00 25.00 25.00 25.00
additional Spraying labor -5.08 -5.08 -5.08 -5.08 -5.08 -5.08
P35110193 Opportunity cost of
application farmer’s capital -0.78 -0.78 -0.78 -0.78 -0.78 -0.78
Incr. Net return to
land (S/ha) 15.15 14.54 14.09 13.68 13.16 13.04
Cotton yield (Kg/ha) 5.90 3.80 4.20 4.75 3.80 3.50
Gross revenue (S/ha) 1.61 1.04 1.15 1.30 1.04 0.95
One Insecticide -28.50 -28.50 -28.50 -28.50 -28.50 -28.50
additional Bt seed 25.00 25.00 25.00 25.00 25.00 25.00
pesticide Spraying labor -5.08 -5.08 -5.08 -5.08 -5.08 -5.08
application Opportunity cost of
farmer’s capital -0.78 -0.78 -0.78 -0.78 -0.78 -0.78
Incr. Net return to
LOW _ land (S/ha) 10.97 9.61 9.72 9.87 9.61 9.53
INPUT Cotton yield (Kg/ha) -8.00 -7.05 -5.60 -5.95 -5.85 -5.1 5
USE Gross revenue (S/ha) -2.18 -1.92 -1.53 -1.62 -1.60 -l .40
Two Insecticide -28.97 -28.97 -28.97 -28.97 -28.97 -28.97
additional Bt Seed 25.00 25.00 25.00 25.00 25.00 25.00
pesticide Spraying labor -5.08 -5.08 -5.08 -5.08 -5.08 -5.08
applications Opportunity cost of ‘
farmer’s capital -0.41 -0.41 -0.41 -0.41 -0.41 -0.41
Incr. Net return to
land (S/ha) 7.27 7.53 7.92 7.83 7.85 8.04
Cotton yield (Kg/ha) -16.85 ~14.80 -l4.25 -12.60 -12.45 -10.75
Three Gross revenue (S/ha) -4.59 -4.04 -3.89 -3.44 -3.40 -2.93
additional Insecticide -29.43 -29.43 -29.43 -29.43 -29.43 -29.43
pesticide Bt seed 25.00 25.00 25.00 25.00 25.00 25.00
applications Spraying labor -5.08 -5.08 -5.08 _ -5.08 -5.08 -5.08
Opportunity cost of
farmer’s capital -0.41 -0.41 -0.41 -0.41 -0.41 -0.41
Incr. Net return to
land (S/ha) 5.32 5.88 6.03 6.48 6.52 6.98

 

-146-

 

 

 

 

Table I-5 (Cont’d).

Cotton yield (Kg/ha) 42.40 38.20 34.40 32.10 29.80 28.85

Gross revenue (Slha) 1 1.56 10.42 9.38 8.75 8.13 7.87

Insecticide -28.03 -28.03 -28.03 -28.03 -28.03 -28.03
No additional Bt seed 25.00 25.00 25.00 25.00 25.00 25.00
pesticide Spraying labor -5.08 -5.08 -5.08 -5.08 -5.08 -5.08
application Opportunity cost of

farmer’s capital -0.41 -0.41 -0.41 -0.41 -0.41 -0.41

Incr. Net return to land

(WE 20.08 18.93 17.89 17.27 16.64 16.38

seed cotton yield

(Kg/ha) 9.45 8.50 7.80 7.60 6.40 6.50

Gross revenue (S/ha) 2.58 2.32 2.13 2.07 1.75 1.77

Insecticide -28.50 -28.50 -28.50 -28.50 -28.50 -28.50
One Bt seed 25.00 25.00 25.00 25.00 25.00 25.00
additional Spraying labor -5.08 -5.08 -5.08 -5.08 -5.08 -5.08
pesticide Opportunity cost of
application farmer’s capital -0.41 -0.41 -0.41 -0.41 -0.41 -0.41

Incr. Net return to land

(S/ha) 1 1.56 11.30 1 1.1 1 1 1.05 10.73 10.75

seed cotton yield

(Kg/ha) -14.25 -12.05 -1 1.60 -10.70 -9.85 -8.70

“K?"- Gross revenue (Slha) -3.89 -3.29 -3.16 -2.92 -2.69 -2.37
32:” Two Insecticide -28.97 -28.97 -28.97 -28.97 -28.97 -28.97

additional Bt seed 25.00 25.00 25.00 25.00 25.00 25.00
pesticide Spraying labor -5.08 -5.08 -5.08 -5.08 -5.08 -5.08
applications Opportunity cost of

farmer’s capital -0.41 -0.41 -0.41 -0.41 -0.41 -0.41

Incr. Net return to land -

(S/ha) 5.56 6.16 6.28 6.53 6.76 7.08

seed cotton yield

(Kg/11a) -30.85 -32.25 -24.90 -23.90 -22.00 -21.75

Gross revenue (S/ha) -8.41 -8.79 ~6.79 -6.52 -6.00 -5.93
Three Insecticide -29.43 -29.43 -29.43 -29.43 -29.43 -29.43
additional Bt seed 25.00 25.00 25.00 25.00 25.00 25.00
pesticide Spraying labor -5.08 -5.08 -5.08 -5.08 -5.08 -5.08
applications Opportunity cost of

farmer’s capital -0.40 —0.40 -0.40 -0.40 -0.40 -0.40

Incr. Net return to land

(tha) 1.50 1.12 3.13 3.40 3.92 3.98

 

Source: Author’s computations based on Strasberg (1997), MADER (2000), and MADER (20018).

-147-

APPENDIX J

Financial and Economic Enterprise Budgets

-148-

 

 

 

 

Nod- Nod- _od. 56. Ed- 86 63% none. .Ent. 2 anm 82.65
36 $6 $6 Ed Etc Ed 33 end mmd 8d Etc Ed 958
Logo. .Efl 8 2.22 oz
36. $6. mod- 36. 25. mod- €58 was 2 E52 82 .85
N _ .3 36m mndm woém mo. :4 om. _ m 8.3 3.3 mad... 3.2 2.3 a. _ m 358 can. 8 2:22 82
8.3 8.? 8.3 8.3 8.3 8.? 8.3 8.3 8.3 8.: 8.3 8.3 one one. on his eon-8:38
34mm 3.3 3.3 RAN 3.3 3.3 3.3 3.3 3.5 3.5 omxm omsm 32:38 622835
.7w .5: Loan— UEE 3.852%
mm; mm; $.— 34 mm; mm; mm; mm: 3.. ww._ mm: 3.. 88:88 28 moo; oz
de nmd de mmd de de and de de de de de 3.533 5.3.3
86 cod 8.9 8.0 86 cod cod cod cod cod ocd cod 6.6653; mm:
cod cod cod cod cod cod cod cod cod cod cod cod noflzta FDA—z.
of: 8.2 8.2 8.2 8.2 8.2 cod cod cod cod cod cod 5:8 poem -30..
cod cod cod cod cod cod cod 86 86 cod cod cod Loafing
cod cod cod cod cod cod 3d wad wad wad wad wad 62283.5
2.3 3.3 cod. 3.2. 34% m _ .§ 2.3 mash no.3 m _ .3. ma. _ m 3. _ o Amp—Ev once-6L $20
:6 :d :d _ _.c :.o __.o :6 _ _.o _ _.o :.o _ 3 2d $25325 5:8 .88
mmm ”.3 m3 «E. can 03 Ohm m _ c 53 $0 m3. mam 3:3: 28..» 8:8 Bum
a h e m v n a h e m w n 8: =3...
3.036.098 .3 .833 «m x83 «53% :e .533 35:53.5 mES. he .25..

 

mwowwsm omtEoEm 30%-Em coxoo Hm

T». 2an

-l49-

 

 

 

 

8...- 8...- 8...- 8...- 8...- 8...- 488. 8...... .5... 6. 532 .oz .8...
..N... X... «a... 3.. 8... 8... R... an... 2.... ?... 8... 8... .28.

.092 .55 on mean. ~02
93- RN. QN- 8a- 8...- 2.?- ?8. on... o. 532 .oz .8... 288:8...
8.?. one. 8.8 9.8 8.8 8.8 8.... 9.2 8.8 an? 2.8 m . .8 .28. 8... 2 .522 32 8.2.82.
8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.8 .8... on 3.5... .2868
8.8 8.2 8.8 8.2 8.8 8.2 8.8 8...? 8.... 8...-.- 88... 8.8 8...-5......- ..e no... .8... 8.... 65
8.. 8.. 8.. 8.. 8.. 8.. 8.. 8.. 8.. 8.. 8.. 8.. 8.2.8:. es. 82.
mm... mm... an... mm... NW... N“... .?... mm... 8... mm... 3.. 2... 88.3.. 6.2% mm:
8.... 8.... 8.2 8.... 8.... 8.... 8... 8... 8... 8... 8... 8... e88 .88 5.2.
8... 8... am... 8... 8... a... ?.e. ?.e. ?.e. ?.-... ?.e. ?.e. 8.2.82. -30..
a . .8 m2..- e..?n 8.8 8.? 3.... on? on. .n 8.8 8. . w 8.... . .8 .28. once-6. no...
..... ..... ..... ..... ..... ..... ..... ..... .2. .2. ..... ..... .aémeocneoeooeooa
.3 a... «.8 mm- efl 8.. e... 38 8.. on. 8.. .8 .28... so... 8:8 88
8...- 8...- 8...- 8...- 8...- 8...- on... Leon. .5... e. E38 32 no...
9.... 9.... S... ?... 8... S... ?... 9.... .2. 8... Q... 8... .82.

.3... .EE o. 2:22 .02
8N- Mc.Nt No. 7 wq _t MON- cw. 7 $53 6:2 9 532 “oz .35 2.232???
8.2 8.8 8.8 2.? 8.8 8.8 m . . .~ 8.3 3.8 83 8. .4 8. . m .28. on... o. 2:32 52 8.2.82.
8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 9.8 .8... on 3.5... .2868
8.8 8.8 8.8 8.8 8.3 8.8 8.8 8.8 8.8 8.8 8.8 8.... $2.38.. ..a .6... nose. on... 25
8.. 8.. 8.. 8.. 8.. w... 8.. 8.. 8.. 8.. 8.. 8.. 8.2.85 es. 82.
mm... NW... $2.. um... ?... 2... 3.. mm... 9.... «m... mm... mm... 8.8.8.. 8.8% mm:
8.2 8.... 8.... 8.2 8.... 8.... 8... 8... 8... 8... 8... 8... =88 .88 ._.:..z.
m2 28 m. .m 2.? m... m...- m-S. 2.2 8.2 3.... mam. 3?. 8.888. -30..
8...... 8.8 8.2- 88 . .8 2.8 3.8 8.8 m . .R an...» 8.2 8.8 .28. 2.8.6- no...
2... .2. ..... ..... ..... ..... ..... ..... ..... ..... ..... ..... .2588 =88 .88
8? .8 c? w ..- m... e? 8? Re 8.. a: a? o? .23.. so... =88 83.

8.280. .-.. 6...:

~150-

 

 

8.... 8...- 8...- 8.... 8...- 8...- 88. .8... 8. :88. 82 .8...
8... 8.. 3... 8.. 8.. 8.. 8... 8.. .~.. 8.. 8.. 8.. .8888.

.88.. 2 “.539. 82
8..- 8..- 8..- 8..- .n..- .8..- 8.8.8.8.... 8.8.82 .8...
8.8 8.8 $8 8.: 8.8 8.8. 8.8 8.... 8.8 8.8 3.8 8...: 8.8.

VGM— On mezaoh ~02
8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.8.38.8 2...... 8.88.8.8
8.8 8.8 8.8 8.8 8.8 8.8 8.8 8.8 8.8 8.8 8.8 8.8 8.8.8.... 8.8.88...
3.. 3.. 3.. 3.. 3.. 3.. 3.. 3.. 3.. 3.. 3.. 3.. 88.8... 8.88.... 88...»...8
8... 8... 8... 8... 8... 8... 8... 8... 8... 8... 8... 8... 88.8.. 8.8.... .5
8.... 8.... 8.... 8.... 8.... 8.... 8.... 8.... 8.... 8.... 8.... 8.... 8.888.. mm:
8.... 8.... 8.8. 8.... 8.... 8.8. 8... 8... 8... 8... 8... 8... .8..8 .88 5.7..
8.. 8..- ...8 8.. 8.8 8..- 8.» 8.. 8.. 8.. 8.. 8.. 88:8. -5...
2.8 2.8 m... m... m... 2.8 8... 8.2 8.... 8.... 8.: 8.2 8.8.88...
8.8. 8.8. 8.8. 8.8. 8.8. 8.8. 8.8. 8.8. 3.8. 3.8. 8.8. 8.?. 8.8. 8.8.8. 8......
z... .2. :8 z... .8 .2. :... .2. z... .8 .8 .8 8.8.8....

5:8 v3...
8. .. 8: m8. 8.... .8. 8... 8.. 8.. m8. 8: 8... 8.... .88...

Eu.» .588 Bu...
.8... 8... 8... 8... 8... 8... .8588. 88. o. 8.8.82 .8...
8... 8.. 8.. 8.. 8.. 8.. 8... 8.. 8... 8.. 8.. 8.. .8588.

.EQ& 0a meaoh “DZ.
8... 8... 8... 2.. 8.. 8.8 8.8.8.8. 8. 53282.8...
8...? 2.... 8.8 8.8 8.... 8...: 8.8 8.8 8.8 8.: 8.8 8.8. 8.8.

was. 9 8:22 .82
8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.? 8.88.8.8 2...... 8.88.8.8
8.8 8.8 8.8 8.8 8.8 8.8 8.8 8.8 8.8 8.8 8.8 8.8 8.8.8.... 88.88...
3.. 3.. 3.. 3.. 3.. 3.. 3.. 3.. 3.. 3.. 3.. 3.. 8.2.8... 8.8.8.. 88...”...
8... 8... 8... 8... 8... 8... 8... 8... 8... 8... 8... 8... 8.8.8.. 88.... z
8.... 8.: 8.... 8.... 8.... 8.... 8.... 8.... 8.: 8.... 8.... 8.... 8.858.. mm:
8.... 88. 8.... 8.... 8.... 8.... 8... 8... 8... 8... 8... 8... 8.8D .88 5.2.
8..- 8... 8.8 8.. 8.. 8..- 88 8.. 8.8 8.8 8.8 8.. 8......8. .8...
8... 8... 8... 8... 8... 8... 8... 8... 8.8 8.8 8.8 8... 88.88...
8.... . 3.8. 8.8. 58. 8.8. 8.8. 8.... . 8.8. 8.8. 8.8. m. .8. 8...: 8...... 8...-.8. 82..
:... .8 .8 .2. .2. z... 2... .8 .2. .2. ..... z... 8.8.8....

cacao B8.
8... 8.. EN. 88. .8. 3?. ?... z: 3.. 88. .8. 8?. 8...»...

 

29S :98“. too..-

 

.......80.

T.- 038-.-

-151-

 

 

 

8...- 8...- 8...- 8...- 8...- 8...- $5.82. ea 3 522 .oz .5...
a... 5.. n... 2.. R. 3.. no... 8.. .N. 2.. 8.. 8.. .35
bond—.53 On 9:32 ~02
«an- 3m- 91. can- .3- 3a- .25 c5. 2 522 .oz .5...
8. . m 3.2 $8 8.? 3.3 3...... on: .o. .o m . .3 $3 8.2 as. . .93. cs. 9 2:22 .oz
8.3 8.3 8.3 8.2 8.3 8% 8.2 8.2 8.: 8.2 8.? 8.2 as 8%. 3 3.5... 22.8.33
3.3 3.2 3.2 33 8.2 an 8.2 8.3 8.2 8.2 8....“ 8.2 5%. c2... ”mfimwmh
3.. m... w... m... a... m... M... a... w... w... m... a... 8.2.82 2a 89. .9;
«no a... a... «no ”no S... m... S... NE. mm... a... 2... 85:2 E29.
3.... 3...: as. 3.: 3.: 3.... 3: 3.: as. 2.... 3.... 3.... ”6.055.. mm:
8.2 8.2 8.2 8.... 8.... 8.2 .o 8... 8... 8... 8... 8... 8:8 88 5.2.
was was was was was was 3 was was was Mas was 55:5. .53:
a... a3 03 a3 a... 3.0 3. SE 2.2 3.2 R... 3.2 8.0.33...
:3. 3.2 8.2: 39.. 2.3. 2.“... SN. 3.3 ”2: ~32 2.3.. No.3. .23. 2.532 320
...o .2. ...c .2. ...o .2. . ..o ...o . I. . ..o .2. ...o @5325. 5:8 E“.m
x. . a. . SN. 8m. Q... ”8. .V. . 3N. m5 «2. 3: NS. .23.. so.» 8:8 .38
.388 I 2.3

-152-

 

 

 

 

 

_oom 3.5m :2825. .5on Mun—<2 «Ea . ooom Bram 3030 20:00 yum—32 A33. .3925 no woman 28:8: 8 9.0523 .350m
8...- 8...- 8...- 8...- o. 2.- o. 2.- ~88. .8... .5... o. 5...... 82 .8...
8... 8... N... S... 8.. 8.. a... 8.. 8.. 2w. 8.. 8.. .8...
.23 EM. 9 mE=.o. 82
.3. 8.1 E..- 9...: 8.9 .2- .25. .8... 9 ea... 82 .8...
8:... 3.2 2.8 m . .2 ~12. 3.8. N2 2...... 3.8 8.8 8.3 8.2 . .25. .8... o. 8.32 82
8.8 8.8 8.8 8.8 8.: 8.? Sn 8.? 8.8 8.8 8.8 8.: S... .8... 8 2...... 82.8.7.8
2.3 8.2. 8.3 on: 2.8 8.8 M...” 2.... 3.: no: 2. . n 2.: .8... 8.... ”flaw“
w... m... w... m... w... w... .2 w... m... w... m... w... 8.2.8... 2... 82. 8:;
N2. N2. N2. N2. N2. 2... 2. N2. 8... a... N2. 8... 8.8.8.. 82......
8.... 8.... 8...... 8.... 8.: 8.... 3.. 8.... 8.... 8.... 8.... 8.... 8.83.... mm:
8.... 8.8. 8.8. 8.2 8.... 8.... .o 8... 8... 8... 8... 82. 8:8 88 5.2.
m3 8.. 8.. .3 was was as 8.. w...» 8.. was was 88...... 45:.
3; 3.... 3.... 3.8 8.8 3; 2.. 8.... 8.... 3...... 8.... 8.... 8.2.8....
8.3. 3.2. a. .3 822 8.8. 2.2... gm. w . .8. 8.3.. o. .8. w _ .8. 8...... .25. 2.88. .85
.2. .2. .2. .2. .2. . 2. .2. .2. .2. .2. . 2. .2. 2.2.8.8.... 8:8 88
8.. SN. 88. a... 8m. .8. 2. mm... :2 2.... mm. a: .23.. 2... 8:8 88
2......8. I 2.5

-153-

 

 

 

 

2... .82 2...... 22 .82 28.... .22. 3.8. 8. :88. 82 .8...
2.8 8.32 23.. 2.2 3. 8. . .22. 283 .83.. 332 8.2... 2...... 382 .2... .8... 8. 888. 82
2.3 2.3 2.3 2.3 2.3 2.3 .3 .3 .3 .3 .3 .3 .288 2.8.8 .8 .80 .880
22... 228 22.. 2.8 228 228 .83... .3... .83... .83... .83... .83... .28. 388 8.8.33 .382
N3... N3... N3... N3... N3... N3... N3... N3... N3... N3... N3... N3... 888.2 88.88...
33.. 23.. 23.. 23.. 23.. 33.. 23.. 2.3.. 23.. 23.. 23.. 3.. 8.8m 8.88.88
22 22 22 22 22 22 22 22 22 22 22 22 883.8 8.8.8.... 88.88...
88 88 88 88 88 8... 8... 88 88 88 88 8... E8. 8. 2.8.... 88.8.2 .22....u8
N3. N3.... N38. N3.... N3.... N38. 3.: 32. 32. 332. 3.: 3.: 88. 2....8. o
.83N .83 .83 8.2 .83N .83 8.2 8.2 8.2 8.2 8.2 8.2 .8...>..8 :3 8.. 8...... 8.... mm:
2.. 2.. 2.. 2.. 2.. 2.. 2.. 2.. 2.. 2.. 2.. 2.. 8.28... 3.8 88.. 5.2.
23.. 23.. 23.. 23.. 23.. 23.. 23.. 23.. 23.. 23.. 23.. 23.. 8.8.8.. 88.8 - 33
2.2 2.2 2.2 2.2 2.2 2.2 2.2 22 2.2 2.2 2.2 2.2 8:8 .88
2... 2... 2... 2... 2... 2... 2.2 2.2 2.2 2.2 2.2 2.2 838.88...
.28. N82. 32. 338. 23.2.N 3.2.2 2.2. 2...: 32.... 82.38. 23.N.N 222 .2... 2.8.8. 88.0
2N... 2N... 2N... 2N... 2N... 2N... 2N... 2N... 2N... 2N... 2N... 2N... .2....8... .8..8 .88
82 .8 .28 2.2 22 .2.. 32 2N8 38 3.2 22 .2.. .28... 3...... .8..8 88
3...: 83. .82. 8.3. 33.3. 33. .22. .83. 8. :88. 82 .8...
38.8 3... . M. 22.. 2...... 82. . 2323. .833 8.... 2.22 .38.. 2.8. 2.2. .28. 38. o. 88.... 82
8. .3 8. .3 8. .3 8. .3 8. .3 8 . .3 28.3 28.3 28.3 8.3 .83 8.3 .28. 2.8.8... .80 .880
828.. 828.. 82.... 828.. 828.. 328.. 8.38 8.38 8.38 8.38 8.38 8.38 .2... 888 28...; 88.2
N3... N3... N3... N3... N3... N3... N3... N3... N3... N3... N3... N3... 88288.8. 88.88...
23.. 23.. 23.. 23.. 23.. 23.. 23.. 23.. 33.. 23.. 33.. 3.. 8.8.8 8.828..
22 22 22 22 N22 N22 22 N22 22 22 22 22 88.88 8.8.8.3.. 8.8.88...
88 8... 88 88 8... 8... 88 88 88 88 8... 88 E8. 8. 88...... 8888.2 8.8.2.8
288. 288. 288. 2.2.. 2...... 288. 8.2. 8.2. 8.2. ...2. ...2. 8. 2. 88. 2:8... oz
3N2 3N2N 3N2 3N2N 3N2 2.2 22N 22N 22N 22N 22N 22N .8...>..8 :3 8.. 88. 8.... mm:
2.. 2.. 2.. 2.. 2.. 2.. 2.. 2.. 2.. 2.. 2.. 2.. 8.28... 8.8 88: 5.2.
23.. 23.. 23.. 23.. 23.. 23.. 23.. 23.. 23.. 23.. 23.. 23.. 8.8.8.. 888m -33
2.2 2.2 2.2 2.2 2.2 2.2 2.2 2.2 22 2.2 22 2.2 .8..8 .88
.8... 8... .8... 8... 8... 8... 8.2 8.2 8.2 8.2 8.2 8.2 8.888...
2.2. 2.2. 83...: 33.... 8.2.2 .32 3.2. 2.8. .22. 2.8. 8. 2.2 8. 22 .22. 888. 8°...
2N... 2N... 2N... 2N... 2N... 2N... 2N... 2N... 2N... 2N... 2N... 2N... 8.2.8.... 8.8.
32 38 38 N82 82 33.. 82 3.8 2.. 28 332 2.. .352... 3.8.2 .888
a 2 e m 3 3. a 2 8 m 3 m

saga—:38 .3 5:8 5 3.3398 .3 5:3 3.3.2.350 2:8. hum: “my...“

 

Sowcsm 8.5.2:”,— o_Eo:oom 2.9.8 5

N-.. 038,—.

-154-

 

 

 

 

.8. 2.. .3.. 8.. ...3 2.3 .2... ...... .. .....8. 82 .8...
83.33 .3... 3.3. .N... ....8 2.2. ...8 ....3 .....3 3.2 8...... ...: . .2... ...... ... 2...... .....m
.2
...3 ...3 ...3 ...3 ...3 ...3 83 ... .3 83 83 83 23 8...... .. .80 .880
.....N. .....N. .....N. .....N. ...... 2..... 2.2. 2.2. 2.8. 2.2. 2.2. 2.2. .2... .88 2......> ......
N3... N3... N3... N3... 2... N3... N3... 2... N3... N3... 2... N3... 22...... 22288...
.3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. 2.. 8.... .2828.
N3... N3... N3... 2... 2... 2... 2... N3... 2... 2... 2... 2... 88...... 22288... 22288...
2.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. 33.. .3.. ...... .2228...
2.. 2.. N... N... N... N... N... 2.. 2.. 2.. 2.. 2.. 82:8 8.22.... 8.....
8.. .... .... .... 8.. 2.... 2.... .... ...... 8.. 8.. 8.. ...... .. 28...... 28......
...... ...... ...... ...... ...... ...... .... .... .... 2.... ...... 2.... 8.8.2.2.. a...
...... ......N ...... ...... ...... ...... 2... 2. . 2 3.... 3.... 33... 3.... .8228. ... 8.. 8.... 8.... 5.2.
2.. 8... 8... 83.. 8... 83.. 83.. 83.. 8... 2.. 2.. 8... 88.8.... ...... 8.... -33
.3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. 8.8.... 8......
....N ....N ....N ....N ....N ....N .... .2.. .... .... .... .... .828 .88
....N .... ....N ....N ....2 ....N .....3 ...... 2..... ...... ...... .....3 8.288...
...... 83.... ...... .3... ... .NN 823 2.2. 8..... ....8. .N....N 8.32 N3.N3N .2... 2:88. 8.....
.N... .N... .N... .N... .N... .N... .2... .N... .N... .N... .2... .N... .2282... .828 .88
2. 3.. 38. .3. . . . 8... 2. ... .... .... .2 .N.. .23.... .22. 8.8.. .88
3.... .... 2.. 8.. .3.. .N.. .2... ...... 2 .....8. ....2 .8...
2.3. ...... .8... 3.... N33... ...... . ...3 . . .3. 8..... N... 8.8 8.2. .2... ...... .. 2.88. ”.82
.... .8
.33 .33 .33 .33 .33 .33 ...3 ...3 ...3 ...3 ...3 ...3 2.8.8 ... 88 .880
2..... 8...... 2..... 8.2... 2..... 22... 38... . 32... . 38... . 38... . 38... . 38... . .2... .88 2......> ......
N3... N3... N3... 2... 2... N3... 2... N3... 2... N3... N3... N3... 29...... 8.288... 2282.8.
.3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. 2.. .3.. 8.8.. 8.888...
N... 2.. N... N... N... N... N... N... 2.. N... N... N... 82:8 8.28.... 882......
8.. .... ...... 8.. .... 8.. .... 8.. 8.. 8.. .... 8.. E... 2 2.8.... .88.... . ...3
.... .... .... .... .... .... ...: ...: ...2. ...: ...: ...: 8.... 8...... 2
.....N 8.2 8.2 8.2 832 83. .....2 ...... ...... ...... 8...... 8...... .8228. ... 8.. 8.... .82.. ...:
8... 83.. 2.. 83.. 8... 8... 2.. 8... 2.. 8... 8... 2.. 88...... .... 8....
.3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. .3.. 8.8.... 8...... ham,”
.... ....N ....N ....N ...2 ....N .... .... .... .... .... .... 82.8 .88
...... ...... ...... ...... ...... ...... ...... ...... ...... ...... ...... ...... 82......
...: 3.... ...: ...: .2... 3.... 2.3 2.3 2.3 2.3 2.3 2.3 22288...
.3.... 2.3.. .N... 8.8. .... .N 2.2. 2.... .8... 2.... ......N 8.8.. ...33... .2... 2:88. 88.0
.N... .N... .N... .N... .N... .N... .2... .N... .2... .N... .2... .N... .2282... .828 .88
. .. .3. 2. 2. 8. 8. ... 83. .... 2. .... .8 .2.... .22.. 8.8.. .88
.....88. N. 2.....

-155-

 

 

 

 

68.6. 68.6. 66... ...: 66... 66... 4656665.... 536262.36.
66.66. 6...6. 366: .866. 66.88 68.688 36.38. 63.63. 68.66. .636. 66.8.8 66.668 356.65.63563632
66.3 66.3 66.3 66.3 66.3 66.3 66.6 66.6 66.6 66.6 66.6 66.6 .3668 86666 .6 .66 .5660
.386. .386. .386. .386. .386. .386. 66. .8. 66. .8. 66..: 66..: 66..: 66. .8. .236. 6.8... 6.63.69 .338
83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 6.68266 66.66866.
63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 6.6.636 8632.663
88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 63.23. 66.6638. 26.56.86...
66.6 66.6 66.6 66.6 66.6 66.6 66.6 66.6 66.6 66.6 86 66.6 23.36.666.66 68668.8 6.66... 6
83.6. 83.6. 83.6. 83.6. 83.6. 83.6. 3.8. 3.: 3.: 3.3.: 3.: 3.: 366.6656. . .666
66.8 66.88 66.88 66.88 66.8 66.88 66. . 6 66. . 6 66. . 6 66. . 6 66. . 6 66. .6 $663666 :6 .6: .666. 62... no
66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 626635 65 68.. 6.6:
83.. 83.. 83.. 83.. 3.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 66.3666 6686.66 .562.
863 8.63 863 863 863 863 863 863 863 863 863 863 66.6.6.2. 4.0:.
8.8 8.8 8.8 8.8 8.8 8.8 8.6 88 8.6 8.6 8.6 8.6 62.8 .686
.688 .688 .688 .688 .688 .688 .688 .688 .688 .688 .688 .688 .6885.
8.6 8.6 8.6 8.6 88 86 66.86 66.86 66.86 66.86 66.86 66.86 66.66866.
6.86 6.6.6 66.666 88.666 66.666 68.63 66.66 83.8.6 8.666 66.66 66.366 6663 .6...6.66622685
88.6 88.6 88.6 88.6 88.6 88.6 886 88.6 88.6 88.6 88.6 88.6 8.66.8.6. 8:8 686
8. . . 68. . 638. 366. 363. 686. 66. . 36.. 668. 686. 63. 6.6. .2862. 6.6... 62.8 686
66.6. 36.6. 8.: 66.: 666. 868 6236.65. 6. 56.2.62 .62..
86.86. 8.86. 68.88. 36.66. 88.88 66.388 36.68. 6366. 86.66. 68.8. 36.668 68.368 8386666226522662
66.3 66.3 66.3 66.3 66.3 66.3 66.6 66.6 66.6 66.6 66.6 66.6 .3668 86.666 .6 .68 .2860
6. .36. 6.36. 6. .36. 6. .36. 6. .36. 6. .36. 66.86. 66.86. 66.86. 66.86. 66.86. 66.86. .6...6. 6.68 6.66.6; .6666.
83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 6.68266 66.66866.
63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 66.666 62.86666
88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 68.5.66 66.65.66. 66.56.63,...
66.6 66.6 66.6 66.6 66.6 66.6 66.6 86 66.6 66.6 66.6 66.6 :5. 26366.... 6.2.65.8 6.666. .
86.6. 86.6. 86.6. 86.6. 86.6. 86.6. 6.: 6.: 6..: 6.: 6.: 6.: 566.6656. _ ...“.66
66.8 66.88 66.88 66.88 66.88 66.88 66.6 66.6 66.6 66.6 66.6 66.6 .8...>..66__26..§.6..62... z
66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 622.63.: 666 666..
83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 66:663.. 69.866 SW66...“
863 863 863 863 863 863 8.63 863 863 863 8.63 863 66.6.6.2. 8.0:.
8.8 8.8 8.8 8.8 8.8 8.8 8.6 8.6 88 8.6 8.6 8.6 69.8 686
.688 .688 .688 .688 .688 .688 .688 .688 .688 .688 .688 .688 6686.5.
66.6 66.6 66.6 66.6 66.6 66.6 66.8 8.8 66.8 8.8 8.8 8.8 66.6686...
...668 .6...6 8.66 63.666 66.866 6.863 38.68 8.86 66.886 66.636 3.886 66.83 .336666632685
88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 €36.85 62.8 363
68. 33: 3.8. 666. .83. 86. 866. 3. .. 8.... 668. 666. 636. .6586: 6.6... 63.8 686

8.6.6888 8-6 6.6.6.8

-156-

 

 

 

66.8 68.6 66.6 68.6 6.6 66.6 86.36.65. 6. 5366.62 .666.
66.86. .363. 66.66. 68.66. :.888 66.388 86.68. 66.3. 66.6. 63.86. 8.8 88.668 866.965.966.26262
66.3 66.3 66.3 66.3 66.3 66.3 66.6 66.6 66.6 66.6 66.6 66.6 .6668 6.6.66.6 .68 66660
66.88. 66.88. 66.88. 66.88. 66.88. 66.88. 86. .6. 86. . 6. 86. . 6. 86. .6. 86. . 6. 86. . 6. .6...6. 6.86 6.66.6.3 .56..
83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 6.666666 66.666666.
63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 66.666
88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 666.366 66.6.6666. 8666.366
66.6 66.6 66.6 66.6 66.6 66.6 8.6 66.6 66.6 66.6 66.6 66.6 .666. 966666.66 62.6.66... 66.66666...
68.6. 68.6. 68.6. 68.6. 68.6. 68.6. 68.8. 68.8. 68.8. 68.8. 68.8. 68.8. 6666.86.66. .28....“666
66.88 66.88 66.88 66.88 66.88 66.88 66.6 66.6 66.6 66.6 66.6 66.6 .66...>..66..666..:666_666... .8
66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.66666. 666 662. 6.6:
83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 66.66666 6666666 8262.
6663 66.63 66.63 66.63 66.63 66.63 66.63 66.63 66.63 66.63 66.63 66.63 66.6562. -85.:
68.68 68.8 68.68 68.68 68.68 68.8 8.6 68.6 68.6 68.6 68.6 68.6 62.8 6666.
.688 .688 .688 .688 .688 .688 .688 .688 .688 .688 .688 .688 668.666..
68.8. 68.8. 68.8. 68.8. 68.8. 68.8. 88.63 88.63 88.63 88.63 88.63 88.63 66.666666.
86.666 36.686 66.636 86.686 66.363 66.863 66.26 66.686 66.636 3.386 66.863 66.663 3.36.666266685
88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 3.36.6666 66666 6666.
36. . 36.. 868. 666. 663. 666. 63. . 368. 688. 886. 663. 886. 66566.. 6.6... 6668 6666

APEOUV N-.. 033.

-157-

 

. Son utsm :26358 8.0on ”En—<2 v5 . coo~ otsm 653° .8on Mmfl<2 . 3a—

..onmabm :0 8606.3 6:265: Eco 6.35:6. ”oasom

 

 

 

66.6 86.6 63.6 6.8 8... 66.. 6:66.666. 6. 56.66.62 .666.
.668. 6.63. 66.66. 88.6. 8.388 .6888 66.38. 38.3. 66.6. 3.66. 66.688 63.888 656.666.626.22662
66.3 66.3 66.3 66.3 66.3 66.3 66.6 66.6 66.6 66.6 66.6 66.6 66.6.66 8.6.66.6 66.666 66660
66.6. 66.6. 66.6. 66.6. 66.6. 66.6. 66.6. 66.6. 66.6. 66.6. 66.6. 66.6. .6...6. 6.666 6.66.66> .568
83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 83.6 6.6666666 66.66666...
63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 63.. 66.666
88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 88.8 66636666266266. 6668.366
66.6 66.6 66.6 66.6 66.6 66.6 66.6 66.6 66.6 66.6 66.6 66.6 :5. 26.66626 66666668 66.666666.
6.: 6.8. 6.: 6.: 6.: 6.: 6.6. 6.6. 6.6. 6.6. 6.6. 6.6. 6666.665... 666%”.66
66.88 66.88 66.88 66.88 66.88 66.88 66.6 66.6 66.6 66.6 66.6 66.6 666.266 =6 66.. 6666. 66.6. 6.8
66.. 66.. 66.. 66.. 66.. 66.. 66.. 66.. 66. . 66.. 66.. 66.. 66.66665 666 666.. 6.6:
83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 83.. 66.66666 6666666 8:62.
66.63 66.63 66.63 66.63 66.63 66.63 6663 66.63 66.63 66.63 66.63 66.63 66.6562. -..oE
68.68 8.8 68.6.8 68.68 8.8 68.68 8.6 68.6 68.6 68.6 68.6 68.6 666.66 6666
.688 .688 .688 .688 .688 .688 .688 .688 .688 .688 .688 .688 668.666.
66.68 66.68 66.68 66.68 66.68 66.68 86.66 86.66 86.66 86.66 86.66 86.66 66.66666...
68.3.6 66.66 66.66 .8886 .66.3 66.663 66.686 66.866 86.666 66.366 .3.6.3 68.663 .6...6.6666366665
88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 88.6 66566666.. 66666 6666
36. . 6.8. 668. 666. 666. 666. 68.. 668. 3. 6. 6.3. 666. 6.8. .2366: 6.6.6 66666 6666

86.6668 8-66.668

-158-

BIBLIOGRAPHY

Aller, H. (1897). Economic Appraising of Transport Projects. Revised ed. Baltimore:
John Hopkins.

Altieri, M. (2001). Genetic Engineering in Agriculture: The Myths, Environmental Risks
and Altematives. Special Report No. 1, Food First, Oakland, California.

Altieri, M. and Rosset, P. (2001). Ten Reasons Why Biotechnology Will Not Ensure
Food Security, Protect The Environment And Reduce Poverty In The Developing
World. http://agbioforum.org[volno34/altieri.htm.

Belli, P, Anderson, J ., Barnum, H, Dixon, J. and Tan, J. (2001). Economic Analysis of
Investment Operations: Analytical Tools and Practical Applications. World Bank
Washington, DC. ISBN 0-8213-4850-7.

 

Bennett, R, Buthelezi, Joseph, T., Ismail, Y. and Morse, S.(2003). Bt cotton, pesticides,
labor and health - A case study of smallholder farmers in the Makhahini Flats,
Republic of South Africa. Outlook on Agg'culture 5555: 123-128.

Bennet, R., Morse, S., Ismail, Y. and Shankar, B. (2003). A drop in the Ocean? The
economic impact of Bt cotton on small-scale producers in South Africa. Report.

Byerlee, D. and Hesse de Polanco, E. (1986). Farmers’ Stepwise Adoption of
Technological Packages: Evidence from the Mexican Altiplano. American
Agicultural Economics Association 68: 519-527.

Munisse, P. (2002) An Overview of the Biosafety Institutional Capacity Building in
Mozambique. In). Proceedings of Workshop on Biosafety Capacity Building In
Eastern and Southern Africa. Held at the Stoc_kolm Environment Institute (SE1),
16-18 October, 2002. Edited by Bhagavan, Maleur and Virgin, Ivar. 2002.

Boughton, D, Tschirley, D, Ofico, O. and Zulu, B. (2002). Cotton Sector Policies and
Performance in Sub-Saharan Africa: Lessons behind the Numbers in Mozambique
and Zambia. FLASH 34E.

Brethour, C and Weersink, A. (2001). An economic evaluation of the environmental
benefits from pesticides reduction. Agricultural Economics 25: 219-226.

-159-

Cabanilla, L, Abdoulaye, T. and Sanders, J. (2003). Economic Cost of Non-adoption of
Bt cotton in West Africa: With Special Reference to Mali. 7‘h ICABR
International Conference on Public Goods and Pigflic Policy for Agricultural
Biotechnology, Revello, Italy, June 29-July 3.

Cabanilla, L and Sanders, J. (2003). New Road Agricultural Modernization: Will Less
Developed Countries Benefit from Biotechnology? (Not published paper)
November 3.

Carrasco-Tauber, C. and Moffitt, L. (1992). Damage Control Econometrics: Functional
Specification and Pesticide Productivity. AmericaL Joumjai of Agricultural
Economics 74: 158-162.

Chambers, R. and Lichtenberg, E. (1994). Econometrics of Pesticide Productivity.
American Journal of Agricultural Economics 76: 407-417.

Chitlhango, S. (2001). Comparacao dos Metodos de Aplicacao de Insecticidas na Cultura
e Alagodao (Gosypium spp.) no Distrito de Morrumbala. Tese de chenciatura na
Universidade Eduardo Mondla_r_i§, Faculdade de Agronomia. Maputo.

Dalsted, N. and Gutierrez, P. (1992). Partial Budgeting. Farm & Ranch Series,
Economics. Colorado State Universitv Coopegltive Extension Service: Forth
Collins, CO, No 3.760,9.

Debertin, D. (1992). Agricultural Production Economics. 2'1d Edition (privately
published).

Edge, J.., Benedict, J., Carroll, J. and Reading, H. (2001). Bollgard Cotton: An
assessment of Global Economic, Environmental, and Social Benefits.
Contemporary Issues. The Journal of Cotton Science, 5: 121-136.

Elbehri, A. and MacDonald, S. (2003). Transgenic cotton and crop productivity: A
general equilibrium analysis for West and Central Africa. Paperjrepared for

msentation at the 6th International Conference on Global Economic Analysis.
The Hague, Netherlands, June 12-14.

 

Elena, M. (2000). Economic Analysis of transgenic vs. conventional cotton in Argentina.
hay/$298intaar/Noticias/EcolAlgodBT.htm.

Environmental Protection Agency. (2001). Bt Cotton Refuge Requirements for the 2001
Growing Season.

http://www.epa.gov/pesticides/biopesticides/pips/bt cotton refiige 2001.htm

 

-l60-

Evenson R. (2004). GMOs: Prospects for Productivity Increase in Developing Countries.
Jgarnal of Agricultural & Food Industrial Organization, vol. 2, article 2.

Falck-Zepeda J., Traxler G. and Nelson R. (2000). Surplus Distribution from the
Introduction of a Biotechnology Innovation. American Journal of Ag’cultural
Economics 82: 360-369.

Feder, G. (1979). Pesticides, Information, and Pest Management under Uncertainty.
American Journal of Agricultural Economics 61: 97-103.

F ernandez-Cornejo, J. and Klotz-Ingram, C. (1998). Economic, Environmental and Policy
Impact Using Genetically Modified Crops for Pest Management. Select presented
paper at the 1998 NEREA Meetinga. Ithaca, NY 22-23 June.

Gould, F. (2000). Testing Bt refuge strategies in the field. News and Views Nature
Biotechnolng 18:3: 266-267

Genissel, A, Augustin, S, Courtin, C, Pilate, G, Lorme, P. and Bourguet, D. (2003). Initial
frequency of alleles conferring resistance to Bacillus thuringiensis poplar in a field
population of Chrysomela tremulae. The Roval Society B 270: 791-797.

Gianessi, L. and Carpenter, J. (1999). Agricultural Biotechnology: Insects control
Benefits. Washington DC: National Center for Food and Am’cultural Policy.
Report. .

 

Gittinger, J. (1982). Economic Analysis of Agricultural Projects, 2nd ed. Baltimore,
Maryland: The John Hopkins University Press.

Glover, D. (2003). Bt cotton: Benefits for poor farmers? Democratizing Biotechnology:
Genetically Modified crops in Developing Countries. BfiefigSeries. Briefing 9.
Brighton. UK. The Institute of Development Studies. ISBN 1 85864 487 9. http:
//www.ids.ca.uk/biotech.

Gouse M, Kirsten, J. and Jenkins, L. (2002). BT Cotton in South Africa: Adoption and

the impact on farm incomes amongst small-scale and large scale farmers.
Working paper: 2002-15. University of Pretoria, South Africa.

Griffin, R, Montgomery J. and Rister M. (1987). Selecting Functional Form in Production
Function Analysis. Western Journal of Agg'cultural Economics 12: 216-227.

Gujarati, D. (1995). Basic Econometrics. 3rd Edition. McGraw-Hill, Inc. ISBN 0-07-
0252149 New York, USA.

-l6l-

Heady, E. and Dillon, J. (1961). Agricultural production functions. Iowa State Press,
Ames, Iowa.

Hilker, J, Black, R. and Hesterman, O. (1987). "Break-Even Analysis for Comparing
Alternative Crops." Extension Bulletin E-202l. MSU Extension, Michigan State
University, East Lansing, MI.

Horstkotte-Wesseler, G, Maredia, M, Byerlee, D. and Alex, G. (2000). Ex-Ante Analysis
in AKIS Projects: Method and Guidelines for Good Practice. World Bank, Rural
Development Family. Report.

Huang, J, Rozelle, S, Pray, C. and Wang, Q. (2002). Plat Biotechnology in China.
Science 295: 674-677. www.sciencemag.org

 

International Foundation for Science. 2004. Data set.

Instituto de Algodao de Mocambique. (2003). Estatisticas de Produao de Algodao na
campanha Agricola 2002/3. Republica de Mocambique.

International Service for the Acquisition of Agri-biotech Applications. (2001). Global
Review of Commercialized Transgenic Cr0ps. ISBN 1-892456-30-3.
http://www.dovlefoundation.org{icsu/ISAAA%20Briefs%2026%20-%202002b.pdf

 

ISAAA. (2003). Pocket K No.5: Documented Benefits of GM crops. m
//www.iszfliorg/kc/Services/html/Pkrihtm.

Ismail, Y, Bennett, R. and Morse, S. (2001). Biotechnology in Afi'ica: The Adoption and
Economic Impacts of Bt Cotton in the Makhatini Flats. Biotechnologv Conference
for Sub-Saharan Africa. Johannesburg, South Afi'ica.

Klotz-Ingram, C, Jan, S, Femandez-Comejo, J. and McBride, W. (1999). Farm-level
Production Effects Related to the Adoption of Genetically Modified Cotton for
Pest Management. AgBioForum 2 (2): 73-84.

Landis, D. (2003). Personal Communication. Professor at Michigan State University,
Department of Crop Systems. East Lansing, Michigan.

Lau L. (1986). Functional Forms in Econometric Model Building. In Z. Griliches and M.

D. Intn'lligator. eds.. Handbook of Econometrica 3: 1515-1564. Amsterdan:
Elsevier.

-l62-

 

 

Lichtenberg, E. and D. Zilberman (1986). The Econometrics of Damage Control: Why
Specification Matters. Americafloaraal of Agricalturg Economics 68: 261-273.

Low, J, Boughton D, Marrule H, Mole, P. and Massingue, J. (2002). O Desempenho do
Salario Minimo: Consideracoes T eoricas e Praticas. MADER/DE. Relatorio de
Pesguisa No. 49P. J aneiro de 2002. Maputo, Republica de Mocambique.

Mabota A. (2002). Determinants of Knowledge of Health Risk and Protective Gear in use
of Pesticides: The Case of cotton Production in Mozambique. MS. Thesis at Ohio
flue Universitv.

McLean, M, Frederick, R, Traynor, Cohen, J. and Komen, J. 2002. A Conceptual
Framework for Implementing Biosafety: Linking Policy, Capacity, and
Regulation. International Service for National Agricultural Research. Briefing
_P_ape_r No.47.

Ministry of Agriculture and Fisheries (1991). A Socio-Economic Survey in the Province
of Nampula: Cotton in the Smallholder Economy. In collaboration with Michigan
State University-Food Security Project. Working Paper 5E. Mozambique.

Ministério de Agricultura e Desenvolvimento Rural. (2000). Inquérito aos Produtores de
Algodao em Nampula. Republica de Mocambique.

MADER. (2001). Inquérito aos Produtores de Algodao em Nampula. Republica de
Mocambique.

MADER. (2001). Cotton Sub-Sector Assessment. Volume 1 (Draft Report)
MADER. (2002). Trabalho de Inquérito Agricola (TIA 2002). Republica de Mocambique.

MADER. (2003). Sistema de Infonncao de Mercados Agricolas (SIMA). Maputo,
Republica de Mocambique.

MADER. (2003). Inquérito aos Inforrnadores Chave em colaboracao com a Universidade
Estadual de Michigan e Projecto de Seguranca Alimentar II, Maputo, Republica
de Mocambique.

MADER. (2004). Private/Public Partnerships to Harness Technology, Markets and
Policies for Accelerated Productivity Growth and Poverty Reduction in
Mozambique. MADER/MSU. A Draft Concept Note in Support of PROAGRI 2
for cwideration bv USAID/Mozambique.

-l63-

Maumbe, B. and Swinton, S. (2003). Hidden health costs of pesticides use in Zimbabwe’s
smallholder cotton growers. Social Science & Medicine.

Morrison, M, Bennett, J, Blarney, R. and Louviere, J. (2002). Choice Modeling and Tests
of Benefit Transfer. American Journal of Agg'cultural Economics 84 (1), pp. 161-
170.

Mullen, J, Norton, G. and Reaves, D. (1997). Economic analysis of environmental
benefits of integrated pest management. Journal of Agricultural Applied
Economics 29: 243-254.

National Technical Information Service. (2003). Country Commercial Guide
Mozambique Year 2003. http://www.buvusa.com or www.cxportgov. or
www.tradeinfo.doc.gov.

Norberto, M. (2003). Personal Communication. Agronomist at the Instituto de Algodao
de Mocambique. Maputo, Mozambique.

Osorio, A. and Tschirley, D. (2003). An Overview of the Cotton Sub-Sector in
Mozambique. Collaborative Research Proiect on Competition a_nc_l Coordination in
Cotton Market Systems in Southern and Eastern Afiica.

Pearce, D. and Howarth, A. (2000). Technical Report on Methodology: Cost Benefit
Analysis and Policy Responses.

Pemsl, D, Waibel, H. and Orphal J. (2004)(in Press). A methodology to assess the
profitability of Bt-cotton: case study results from the state of Karatanaka, India.
Crop Protection.

Pitoro, R, Govene, O, Marrule, H, Tschirley, D. and Boughton, D. (2002). Cotton Sub-
sector performance at farm level in the province of Nampula: Current Situation
and Improvement Perspectives (Version in Portuguese). Ministry of Agriculture
and Rural Development, Maputo-Mozambique. Research Paper 47P.

Pray, C. and Ma, D. (2001). Impact of Bt Cotton in China. World Development 29 (5):
813-825.

Puri, S, Murthy, K. and Sharrna, O. (1999). Handbook of cotton in India. Indian society
for Cotton Irrmrovement (ISCI): 233-245, Mumbai, India.

Qaim, M. and Zilberman, D. (2003). Genetically Modified Crops can have Significant
Yield Effects in Developing Countries. Science 299: 900-902.
www.sciencemag.org

 

-164-

Qaim, M. and de Janvry, A. (2003). Genetically Modified Crops, Corporate Pricing
System, and Farmer’s Adoption: The Case of Bt cotton in Argentina. American
Joum_al of Agricultural Economics 85 (4): 814-828.

Republic of Mozambique. (2003). Review of the Economic and Social Plan for 2003.
March 2004. http:/l www.worldbaLk.org/files/BPES 2003 FMI.pdf

Rodrigues, N. (2002). Manual da Cultura de Algodao. AGRIMO. II Edicao, Junho.
Mocambique.

Roush, R. (1998). Two-toxin Strategies for Management of Insecticidal Transgenic crops:
can Pyramiding succeed where Pesticide mixtures have not? The Royal Socieg
London B 353: 1777-1786.

Schmitz, A. (2004). Controversies Over The Adoption of Genetically Modified
Organisms: An overview of the Special Issue. Journal of Ag’cultural & Food
Industrgl Organization, vol. 2, article 1.

Shankar, B. and Thirtle, C. (2003). Pesticide Productivity and Transgenic Cotton
Technology: The South Afiican Smallholder Case. _P_aper presented L 8th
mposium of Intematioaafilouncil for Agricultural Biotechnology ReseaLcla,
Revelogalv. June 30‘h -Julv 3'“.

Silvie, P, Deguine, J, Nibouche, S, Michel, B. and Vaissayre, M. (2001). Potential of
threshold-based interventions for cotton pest control by small farmers in West
Afiica. Crop Protection 20: 297-301. Elsevier Science Ltd.

Strasberg, P. (1997). Smallholder Cash-Cropping, Food-Cropping and Food Security in
Northern Mozambique. Research Paper. Reprint of PhD. Dissefirtion at MSU-
Agricultural Economics.

Swinton, S. (1998). Less is More: Why Agro-chemicals Use Will Decline in
Industrialized Countries. In Economics of Aggro-chemicals. An International
Overview of Use PJatterns. Technical and Institutional DetermingtsaPolicieflg
Perspectives. Ashgate -Brookfield USA: 359-372.

Tabashnik, B, Lin, Y, Malvar, T, Heckel, D, Masson, L. and Ferre, J. (1998). Insect
resistance to Bacillus thurigiensis: Uniform or diverse? The Roval Societv London
B 353: 1751-1756.

Tabashnik, B, Dennehy, T, Carriere, Y, Liu, Y, Meyer, S, Patin, A, Sims, M. and Ellers-
Kirk, C. (2003). Resistance Management: Slowing Pest Adaptation to Transgenic
Crops. Acta Agric. Scand., Sect. B, Soil and Plant Sci. Supplementum 1: 51-56.

 

-165-

 

Taylor & Francis.

Tothova, M. and Oehmke, J. (2004). Genetically Modified Food Standards as Trade
Barriers: Harmonization, Compromise, and Sub-Global Agreements. Journal of

Agricultural & Food Industrial Organization, vol. 2, article 5.

Traxler, G, Avila, S, Zepeda, J. and Arellano, J. (2001). Transgenic cotton in Mexico:
economic and environmental impact. Paper prepared for the 5th International
Conference, Biotechnologya Science and Modern Agricultare: A new Industry at
the Dawn of the Centug, Revello, Italy.

Traxler, Greg. (2004). The Economic Impact of Biotechnology-Based Technological
Innovations. ESA Working Paper No. 04-08. www.fao.org/es/esa.

Tschirley D. and Zulu, B. 2003. Zambian Cotton in Agricultural Context: Performance
under Liberalization and Future Challenges. Food Security Reasearch Project-
Zambia. Policy Synthesis No.7.

United Nations Environmenta Programme-Global Environment Facility. 2002.
Biodiversity Enabling Activities. _S_ub-Proiect Document. GF/2716-01-4319.
GF/2716-02-4. Mozambique.

United States Department of Agriculture. (2001). Bt Cotton & Management of the
Tobacco Budworm-Bollworrn Complex. Agg'cultural Research Service ARS-154.
January.

Verbeek, Mamo (2000). A Guide to Modern Econometrics. John Wiley & Sons, LTD.
New York. ISBN 0-471-89982-8.

Walker, T. (2003). Field Trip to Nampula. Department of Policy Analysis. Ministry of
Agriculture and Rural Development. Maputo, Mozambique.

Wally, G. (2004). Personal Communication. Biotechnology Regulatory Manager. South
Afiica.

Woodridge J. (2000). Introductory Econometrics: A Modern Approach, South Western
College Publication.

World Bank Institute. (2003). Country-level Economic Performance Data set.

Zilberman, D, Ameden, H, Graff, G. and Qaim, M. (2004). Agricultural Biotechnology:
Productivicty, Biodiversity, and Intellectual Property Rights. Journal of

Agg'cultural & Food Industrial Organization, vol. 2, article 3.

-166-

   

u'[iiiljir‘ljjijjirim: