Fﬁéﬁarﬁmﬂ 9
k3: A. n13.11-
Au: dummy.
1 .3 I
an... it...
2:39. 3...»?pr
, man»... 2%,. r
. . :96» 50......
.. Lu}. 7 $1.34...
:33..." 3%. r u .2
. 3,5! I... , J...
. hit»; . h.6.......
, dﬂﬁiﬁ hummmﬁir
. . . .39nﬁxuujﬁeu a
. . .xiu........ .. . . as
. . . u «3.11;
.. .
“at...

r.
is
r... ......hm..m«upﬁﬁ. ,
- ..
Em. quagmiﬁ
.. :. ..
13...;
.

L... o)...
. xi E. . x:
.5:

.. Er
n. a...
.2...-
t

if
- L“

f

 

 

. . ﬂ '1.»
.. JFHH‘i
11.6...
'3’!
.4...

 

3.9..

x15 .. I

5?.

 

a - ... Egvumm

. z 3 t......;5....~...m.1n...\m:+u
.IJBI . . .. . . 34%...» gnaw. . g .3 ﬁr}: .
i . . . .
1.. 3.... . .. ..

“ES‘” LIBRARY

“‘ Michigan State
5.3) .. University
This is to certify that the

dissertation entitled

ESSAYS ON THE ECONOMIC EVALUATION OF
INTEGRATED PEST MANAGEMENT EXTENSION IN
NICARAGUA

presented by

RICARDO ANTONIO LABARTA-CHAVARRI

has been accepted towards fulﬁllment
of the requirements for the

PhD. degree in Agricultural Economics

 

 

} //
/ ."f /7 [2/ . "/I‘
ss‘br’s ' aure

Major Profe

 

I)“ 7.91 7.700%:

 

Date

MSU is an Afﬁnnative Action/Equal Opportunity Institution

 

..--.--u—-—~-- --

 

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

 

030? 43017199

 

 

 

 

 

 

 

 

 

 

 

 

 

 

2/05 p:/CIRCIDaIeDue.indd-p.1

ESSAYS ON THE ECONOMIC EVALUATION OF INTEGRATED PEST
MANAGEMENT EXTENSION IN NICARAGUA

By

Ricardo Antonio Labarta-Chévarri

A DISSERTATION

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

DOCTOR OF PHILOSOPHY
Department of Agricultural Economics

2005

€55

CV2

115:

C l'(

im

 

Cl

ABSTRACT
ESSAYS ON THE ECONOMIC EVALUATION OF INTEGRATED PEST
MANAGEMENT EXTENSION IN NICARAGUA
By

Ricardo Antonio Labarta Chavarri

This dissertation assesses the economic impact of delivering integrated pest
management (1PM) training through farmer ﬁeld schools (FFS) in Nicaragua. The three
essays are based on a cross sectional survey of 436 bean growers. The ﬁrst essay presents
evidence that prior experience with the environmental and health (E&H) effects of
pesticides affect household decisions about pest management, but only when farmers are
using toxic or highly toxic pesticides. When facing high health risks, farmers also tend to
use protective strategies like hiring pesticide applicators.

The second essay evaluates FFS impacts on households’ pest management, bean
crop income and E&H outcomes. The E&H outcomes are represented by changes in the
incidence of acute health symptoms and changes in the observed level of beneﬁcial
insects. One analytical complication is that FFS participants are not randomly selected
from the population of farmers; they tend to be better managers on average. Results show
that failing to correct econometrically for the endogeneity effects of this nonrandom
selection associated with FF S participation can exaggerate the impacts of FFS. After
correction for endogeneity this dissertation ﬁnds FFS performance to be inferior to other
1PM training programs, at inducing IPM adoption, raising net income from beans and

improving E&H outcomes.

The third essay explores the causes of poor FFS performance in Nicaragua by
evaluating whether FFS impacts are inﬂuenced by the characteristics of the speciﬁc non-
governmental organizations (NGOS) that implemented F FS. Two categories of NGO-
speciﬁc effects make a difference, their institutional characteristics and the characteristics
of their F FS on-farm research. Effective NGOS conducting successful F FS have many
years of experience working with the targeted farmers, they have more extensionists, and
most of their extensionists are trained in IPM (not necessary in FFS). When NGOS
deviate from the original FFS focus on IPM training and shift the FF S focus to other
institutional interests like promotion of credit programs, the FFS become less effective at
disseminating IPM and improving farm beneﬁts. Given the participatory research focus
of F FS, the outcomes of on-farm FFS ﬁeld experiments were also important. Farmers
who observed higher yields or net incomes in the IPM plot compared with the
conventional pest control plot were more likely to adopt IPM and reduce pesticide use.
However, most comparisons between an IPM plot and a conventional plot resulted in

lower yields and lower net revenues in the IPM plots.

© Copyright by
RICARDO ANTONIO LABARTA-CHAVARRI
2005

Con mucho Amor a:
Malena, Camila y J osé Maria, mi pequei’ia gran familia

Emilio y Ana, mis padres

ACKNOWLEDGEMENTS

I am very grateful to Scott Swinton, my major professor, assistantship supervisor,
thesis supervisor and best friend, for his support and guidance during my whole graduate
program. He has largely contributed to make the last four years the most productive and
enjoyable in my professional career. His critically and constructive suggestions have
always challenged me and make me give my best efforts. I am also grateful to Scott’s
wife Sylvia Morse and their children for being supportive and a good example for my
family. My apologizes to them for sometimes demanding Scott’s family time to deal with
my academic requirements. I also thank Dr. Sandra Batie, Dr. James Oehmke and Dr.
Jeffrey Wooldridge for serving in my committee and for providing me valuable feedback
on my dissertation.

This research would not have been possible without the support of the “Programa
para el Manejo Integrado de Plagas en America Centr ” (PROMIPAC) that is run by the
Escuela Agricola Panamericana “El Zamorano” and funded by the Swiss Agency for
Development and Cooperation (COSUDE). They ﬁmded a large part of the ﬁeld work
and provided me feedback during the dissertation writing. I am grateful to Dr. Alfredo
Rueda for his sustained interest in my research and for creating an enabling environment
for the ﬁeld work. I also thank Julio LOpez, Orlando Céceres, Francis Porras, Freddy
Soza, Felipe Pilarte, Harold Arguello, Yordana Valenzuela, Zenia Flores, Maritza
Canales, Edgar and Alfredo Castro from PROMIPAC-Nicaragua for their support during

my stay in Nicaragua.

vi

 

Dr

VI"

th

 

 

I am extremely grateful to the team of enumerators: Oscar Gutierrez, Rafael
LOpez, Marisela Benavides, Marisol Altamirano and Meylin Marin. Their ﬁeld
experience, dedication, sense of humor and especially their ﬁ'iendship made it possible to
complete good quality ﬁeld work. I also thank the following institutions in Nicaragua:
AsociaciOn para la Diversiﬁcacién y Desarrollo Agricola (ADDAC), Caritas Diocesana
de Jinotega, Centro de Estudios de Eco-Desarrollo para el TrOpico (CECOTROPIC),
Empresa de Servicios Técnicos Agropecuarios (ESETECA), Fundacién para la
Investigacién y el Desarrollo Rural (FIDER), Organizacién para el Desarrollo Municipal
(ODESAR) and UniOn Nacional de Agricultores y Ganaderos (UNAG-Esteli). Especially
I thank Orontes Rodriguez, Marco Zeledén and Marco Diaz. I express my appreciation
also to the Nicaraguan bean growers that kindly participated in this research and provided
the information for this dissertation.

I also wish to thank for ﬁnancial support the US. Agency for International
Development’s Bean-COWpea Collaborative Research Support Program (CRSP) and the
Department of Agricultural Economics for funding my assistantships and part of the ﬁeld
work, and the College of Agriculture and Natural Resources and the Graduate School for
the Dissertation Completion Fellowship.

I would like to express also my gratitude to my former supervisors in the research
centers that I worked before joining MSU. They were highly supportive and always
encouraged me to continue graduate studies. I especially thank Sam Fujisaka, Doug
White and Peter Kerridge at CIAT, John Weber and Dale Bandy at ICRAF, and Jorge

Recharte, Tom Walker and Elias Mujica at CIP.

vii

I am also grateful to my professors in the Agricultural Economics department
from whom I learned a lot about how applied economics can be useful and help to solve
problems in the developing world. I will also cherish the good environment that I had in
Cook Hall with many friends and a diversity of graduate students. I specially thank
Fernando Balsevich, Laura Donnet, Lourdes Martinez, Meeta Punjabi, David Mather,
Abdoul Murekesi, Antony Chapoto, Wei Zhang, Lilian Kirimi, Mary Mathenge, Lesiba
Bopape, Kirimi Sindi, Ricardo Hernandez, Gustavo Puente, Anne Cullen, Orlando Mejia,
Byron Reyes, Daniel Ortega, Juan Peﬁaherrera, Francisco Arellano, Dan Mooney,
Valeria Sénchez and Cristébal Aguilar.

AS usual I reserve my last words to express my gratitude and my love to Malena,
Camila and J osé Maria, my every day inspiration, for always providing me support and
for understanding my absence from home during many hours during my graduate
program: Los Amo mucho! Also, I would like to express my gratitude and debt to Emilio
and Ana, my parents for having showed me this path and having supported me always

during the whole process: Misién cumplida queridos viejos!

viii

TABLE OF CONTENTS

LIST OF TABLES .................................................................................. x
LIST OF ABBREVIATIONS ..................................................................... xii
INTRODUCTION .................................................................................... 1

ESSAY 1: DO ENVIRONMENTAL AND HEALTH CONCERNS INFLUENCE
FARMERS’ INPUT DECISIONS? EVIDENCE FROM PESTICIDE USE AND

ADOPTION OF INTEGRATED PEST MANAGEMENT IN NICARAGUA .............. 7
1. Introduction ................................................................................... 8
2. Pesticide use, environmental and health effects, and the adoption of Integrated

Pest Management in Nicaragua ............................................................ 9

3. The analytical framework ............................................................... 11
3.1. Farmers’ awareness and experience of E & H effects ...................... 11

3.2. Chemical and non-chemical input decisions among small farmers ....... 12

4. Model empirical implementation ......................................................... 16
4.1. The separability test ............................................................. 16

4.2. The data ............................................................................ 17

4.3. The econometric estimation .................................................... 18

4.3.1. Potential econometric problems .................................. 18

4.3.2. The econometric speciﬁcation .................................... 19

4.3.3. The dependent variables .......................................... 19

4.3.4. The explanatory variables ......................................... 20

5. Results ....................................................................................... 22
5.1. Farmer awareness of pesticide-ascribed health and environmental
effects ............................................................................. 22

5.2. Farmers’ direct experience with environmental and health effects ....... 23

5.3. The inﬂuence of E & H effects on household production decisions......25

5.3.1. Demand for pesticides ........................................... 25

5.3.2. IPM adoption ...................................................... 28

5.3.3. Labor demand ..................................................... 31

6. Conclusions ............................................................................... 33
References ..................................................................................... 35

ESSAY 2. DOES PARTICIPATION IN FARMER FIELD SCHOOLS IMPROVE
GRADUATES’ FARM, HEALTH AND ENVIRONMENTAL OUTCOMES? THE

CASE OF NICARAGUAN BEAN GROWERS ............................................ 38
1. Introduction ............................................................................... 39
2. Farmer ﬁeld schools and the delivery of IPM ....................................... 41
3. FFS experience among Nicaraguan bean growers ................................... 43

ix

4. The analytical framework ............................................................... 45

5. Empirical analysis ....................................................................... 49
5.1. Data ................................................................................ 49

5.2. Potential econometric problems ............................................... 50

5.3. The econometric speciﬁcation ................................................ 51
5.3.1. The dependent variables .......................................... 53

5.3.2. The explanatory variables ......................................... 54

6. Results .................................................................................... 56
7. Conclusions .............................................................................. 65
References .................................................................................... 69

ESSAY 3. MULTI-INSTITUITIONAL IMPLMENTATION OF FARMER FIELD
SCHOOLS AMONG NICARAGUAN BEAN GROWERS. DO DIFFERENT NGOS
PERFORM DIFERENTLY? ......................................................................................... 73
1. Introduction ............................................................................. 74
2. NGOS and their participation as extension providers in developing countries.75
3. The extension service in Nicaragua and the delivery of IPM through a multi-

institutional FFS project ............................................................... 78

4. The analytical framework .............................................................. 84
4.1. The sample design and data collection ........................................ 85

4.2. The econometric estimation .................................................... 86

4.2.1. Potential econometric problems .................................... 86

4.2.2. The econometric speciﬁcation ....................................... 88

4.2.3. The dependent variables ............................................. 89

4.2.4. The explanatory variables ............................................ 90

5. Results ..................................................................................... 92

6. Conclusions ............................................................................... 102
References .................................................................................. 105
CONCLUSIONS ................................................................................ 110
APPENDIX A]. Household survey ........................................................ 113
APPENDIX A.2. Map of surveyed areas in Nicaragua .................................. 126
APPENDIX.A.3. Weighting scheme to correct the sample stratiﬁcation ............. 128
APPENDIX A.4. Production and Socioeconomic Variables used in all regression
modelsl30
APPENDIX A.5. Potential measurement error in Essay 1’s environmental and health
variables ........................................................................................ 133
APPENDIX A.6. Acute health symptom frequencies used in Essay 1 ................ 135
APPENDIX A.7. Complete regression results of essay 1 .............................. 137
APPENDIX A.8. Complete regression results of essay 2 .............................. 144
APPENDIX A.9. Complete regression results of essay 3 .............................. 158

LIST OF TABLES

ESSAY 1
Table 1.1. Summary of variable statistics. 436 Nicaraguan bean growers, 2003-04......24

Table 1.2. The jointly effect of E & H variables on household input decisions, F—test

for 436 Nicaraguan bean growers, 2003-04 ................................... 26
Table 1.3. Inﬂuence of E & H effects on pesticide use, Survey least squares regression

results for 436 Nicaraguan bean growers, 2003-04 ................................... 27
Table 1.4. Inﬂuence of E & H effects on the adoption of different IPM activities. Survey

multinomial logit results for 436 Nicaraguan bean growers, 2003-O4 ............. 30
Table 1.5. Inﬂuence of E & H effects on labor demand for pesticide application. Linear

survey regression results for 436 Nicaraguan bean growers, 2003-04 ............. 32
ESSAY 2

Table 2.1. Differences in pesticide use, adoption of IPM practices and bean yields and net
revenues among ﬁve groups of Nicaraguan bean growers during 2003 bean
season ....................................................................................... 57

Table 2.2. Descriptive statistics of selected variables used in the survey least squares for
436 Nicaraguan bean growers, 2003-04 ................................................ 59

Table 2.3. The effect of FFS and other IPM training on IPM adoption. Survey least
squares regression results for 436 Nicaraguan bean growers, 2003-04 .......... 61

Table 2.4. The effect of FF S and other IPM training on IPM adoption. Survey least
squares regression results for 436 Nicaraguan bean growers, 2003-04 .......... 62

Table 2.5. The effect of FF S and other IPM training on bean yields and net revenue.

Survey least squares regression results for 436 Nicaraguan bean growers, 2003-

Table 2. 6. The effect of FFS and other IPM training on household health outcomes.
Survey ordered probit regression results for 436 Nicaraguan bean growers, 2003-

Table 2.7. The effect of FFS and other IPM training on environmental outcomes. Survey
ordered probit and probit regression results for 436 Nicaraguan bean growers,
2003-04 ...................................................................................... 65

xi

ESSAY 3
Table 3.1. Description of NGOS working on F F S implementation in Nicaragua .......... 81
Table 3.2. Curricula and other activities developed by each bean FFS in Nicaragua. ....83.

Table 3.3 Descriptive statistics of selected variables used in the survey least squares for
436 Nicaraguan bean growers, 2003-04 ................................................ 93

Table 3.4. Husman test and joint signiﬁcance F-test of aggregate FF S treatment effect on
household pesticide demand, adoption of IPM practices and bean net revenue
models for 436 Nicaraguan bean growers, 2003-04 .................................. 94

Table 3.5. FF S and NGO characteristics effects on adoption of IPM practices. Least
squares survey regression results 436 Nicaraguan bean growers, 2003-04 ........ 95

Table 3.6. F FS and NGO characteristics effect on pesticide demand. Least squares survey
regression results for 436 Nicaraguan bean growers, 2003-04 ...................... 97

Table 3.7. FFS and NGO characteristics effect on bean net revenues. Least squares
survey regression results for 436 Nicaraguan bean growers, 2003-04 ............. 99

xii

ADDAC:

APP:

BCN:

CARITAS:

CECOTROPIC:

E&H:

ESETECA:

FFS:

FIDER:

INEC:

INTA:

IPM:

MAGFOR:

MARENA:

NGO:

ODESAR:

PAHO:

PANNA:

PROMIPAC:

T&V:

UNAG:

LIST OF ABBREVIATIONS

Asociacién para la Diversiﬁcacién y Desarrollo Agricola Comunal
Acute Pesticide Poisoning

Banco Central de Nicaragua

Caritas Diocesana

Centro de Estudios de Eco-desarrollo para el Trépico
Environmental and Health

Empresa de Servicios Técnicos Agropecuarios

Farmer Field Schools

Fundacién para la Investigacién y el Desarrollo Rural

Instituto Nacional de Estadisticas y Censos

Instituto Nicaragﬁense de Tecnologia Agropecuaria

Integrated Pest Management

Ministerio de Agricultura, Ganaderia y Forestales

Ministerio de Ambiente y Recursos Naturales
Non-Governmental Organization

Organizacién para el Desarrollo Comunal

Pan American Health Organization

Pesticide Action Network North America

Programa para el Manejo Integrado de Plagas de América Central
Training and Visit extension method

UniOn Nacional de Agricultores y Ganaderos

xiii

INTRODUCTION

The new conceptualization of agricultural extension services in developing
countries has introduced many changes in the delivery of agricultural technologies
(Qamar 2002, Suleiman & Hall 2004). The operational capacity of public extension has
been dramatically reduced aﬁer following criticism for insufﬁcient impacts and the recent
wave of structural adjustment programs (Farrington 1994). New extension providers and
new methods for delivering agricultural technologies have emerged, producing
signiﬁcant changes to the traditional means of delivering extension services.

A diversity of decentralized private providers and non-governmental
organizations (N G0) are replacing the large and centralized public institutions in charge
of delivering technologies (Rivera 2001, World Bank 2002). Also new approaches to
extension have been proposed in order to improve the adoption of agricultural
technologies. These new approaches use participatory techniques and favor a more active
participation of farmers in the search for combining scientiﬁc knowledge with farmers’
experience and interests (Kenmore 2002). Measuring the impacts of the “new” extension
system in developing countries has become a major challenge for agricultural economists.

This dissertation contributes to the economic impact assessment of this “new”
extension system by analyzing the delivery of integrated pest management (IPM) through
the farmer ﬁeld schools (FFS) extension approach that has received increasing support
from development agencies as the preferred means for disseminating IPM (Feder et al
2004). IPM is a knowledge-intensive set of management practices that aim to reduce

environmental and health (E & H) risks related to pesticide use by using chemical pest

control methods only when strictly necessary and by offering alternative non-chemical
pest control methods. FF S is an intensive IPM training program that incorporates
farmers’ priorities and a “learning by doing” approach to skill development among
farmers. FFS proponents expect it to increase the rate of IPM adoption compared to
traditional extension approaches, notably the training and visit system.

This dissertation uses the recent implementation of FF S in Nicaragua as a case
study and addresses different aspects of the impact of the delivery of IPM through FFS
among bean growers. Nicaragua is one of the poorest and least developed countries in
Latin America; two thirds of its rural inhabitants are poor and more than 25% are
extremely poor (World Bank 2003). Agriculture remains the main productive activity and
is responsible for 29% of the total gross domestic product (BCN 1998). Small farmers
dominate the Nicaraguan agriculture, with 48% of farmers having less than 7 hectares of
land (INEC 2001). Most of the small farm households are led by males (84%) but
women’s contributions to household income generation have been increasing, especially
with a greater participation on off-farm activities. However, the gap between women’s
and men’s earnings remains large, with women earning on average 44% of the male wage
(IN EC 2001). During 2003 the average rural wage across Nicaragua ranged between 20
and 40 cordobas per day (between 1.25 and 2.50 US$).

Few crops are responsible for most of the agricultural production in Nicaragua.
Maize and beans are by far the most widely planted crops at 392,525 and 229,215
hectares, respectively. Other major crops include coffee (129,910 ha), sorghum (81,145
ha), plantain (63,491 ha) and rice (52,495 ha). Most of these crops are grown in different

regions according to where the agro climatic conditions are most favorable. The western

departments of J inotega and Matagalpa are the main areas for producing maize, beans
and coffee, Sorghum are mainly grown in the coastal departments of Leén and Managua,
plantain in the RegiOn Auténoma del Atlantico and the department of Rivas, and rice in
the central departments of Granada and Matagalpa (IN EC 2001).

Beans are cultivated by 57% of Nicaraguan farmers, with 40% of these farmers
owning less than 7 hectares of land. Among the grain crops (maize, beans and sorghum)
that constitute around 40% of the total value of agricultural production in Nicaragua, only
maize production is more important among small farmers (IN EC 2001). Although
Jinotega and Matagalpa are the main areas of bean production, the crop is also important
in Regién Auténoma del Atlantico Sur (40,371 ha), Regién AutOnoma del Atlantico
Norte (34,347 ha) and the department of Esteli (17,588 ha). In all these regions beans are
mainly considered a subsistence crop; however, bean sales can also constitute an
important source of income among Nicaraguan producers.

Nicaragua has two major crop farming seasons known as primera (May- August)
and postrera (September-December). A third minor season known as apante (J anuary-
April) is limited to certain areas that have longer rainy seasons than the traditional
postrera. Beans are mainly produced during the postrera season ( 53%) and in less
quantity during the primera (30%) and apante seasons (17%) (MAGFOR 2003). Average
bean yields are around 650 kg/ha (MAGFOR 2003) and are directly affected by insects
like whiteﬂy and bean pod weevil, and by diseases like web slight and angular leaf spot.
Several types of weeds are also a big problem among bean growers. Given the severe
damage that these insect pests, diseases and weeds can produce in beans, pest

management is important to bean producers. Most Nicaraguan bean growers rely on

chemical pest and weed control, with 75% using insecticides and 60% using herbicides.
The main insecticides used are methamidophos, malathion, methyl-parathion and
chlorpyrifos, while the main herbicides are paraquat, 2,4-D and glyphosate. Non-
chemical control options are not widely diffused due to limited farmer access to technical
assistance and credit programs.

Agricultural extension has been the responsibility of the public sector, especially
the Instituto Nicaragtiense de Tecnologia Agropecuaria (INTA). However during recent
years, a structural adjustment process led by the central government for reducing ﬁscal
deﬁcits has led to reduce public services and encouraged the entrance of several non-
governmental organizations (N ODS) and private institutions as major extension
providers. By 2001 , only 12% of Nicaraguan farmers had access to technical assistance.
Of this total, 40% were served by NGOS, only 28% by public entities and 11% by private
extension providers (the rest by different small types of organizations) (INEC 2001).

In this context, the new extension providers have increased their participation in
delivering agricultural technologies and in new extension approaches. One of these cases
has been the delivery of IPM that has for many years produced a low adoption rate. Many
NGOS are increasingly assuming the responsibility of conducting IPM training programs
and most recently a group of them have committed to implement the IPM training trough
the novel FFS approach. The three essays of this dissertation analyze in detail the
implementation of 13 FF S among bean growers in Nicaruagua.

This dissertation is based on data from a cross sectional survey of 436 bean

growers that was carefully designed using a double stratiﬁcation (Deaton 1997) to

compare the effects of alternative IPM training programs (FF S and others) as well as to
include farmers not participating in IPM training programs.

The ﬁrst essay aims to test the underlying assumption of many of IPM delivery
programs that the E&H effects of pesticide use will encourage farmers to reduce pesticide
use and seek alternative pest control methods. To do so, this essay evaluates
econometrically whether farmers’ awareness of and experience with acute pesticide
poisoning symptoms and beneﬁcial insects (E&H effects associated with IPM adoption)
affect their decisions about pesticide use, adoption of non-chemical pest control methods
and the hiring of laborers to spray pesticides.

The second essay evaluates the F FS implementation in Nicaragua and measures
econometrically the impacts of FFS in four dimensions: 1) the adoption of IPM practices
(insect scouting, botanical insecticides and insect yellow traps), 2) toxicity of pesticides
used, 3) bean yields, and 4) net revenues. This essay also incorporates F FS impacts on
E&H outcomes represented by changes in the incidence of acute health symptoms and
changes in the level of observed beneﬁcial insects. The evaluation controls for potential
endogenity due to non-random placement FFS and farmer participation in IPM training
programs.

The third essay, which is motivated by results from essay 2, explores whether the
speciﬁc characteristics of the institutions involved in implementing F F8 in Nicaragua
have inﬂuenced the FPS impacts. Controlling for endogeniety and for the clustered and
stratiﬁed sample design, this essay uses econometric techniques to explore how the FFS
treatment effect is inﬂuenced by NGO institutional focus, resource capacity, experience

and expertise.

 

 

 

References
BCN (1998). Banco Central de Nicaragua. Cuentas Nacionales de Nicaragua. Managua.

Deaton, A. (1997). The Analysis of Household Surveys. A Microeconometric Approach to
Development Policy. The John Hopkins University Press, Baltimore and London.

F eder, G., R. Murgai and J. Quizon (2004a). Sending farmers back to school: The impact
of farmer ﬁeld schools in Indonesia. Review of Agricultural Economics. 26(1):
45-62.

Farrington, J. (1994). Public sector agricultural extension: Is there life after structural
adjustment? Natural Resource Perspectives 2. ODI, London.

IN EC (2001). Instituto Nacional de Estadisticas y Censos. III Censo Nacional
Agropecuaria. Managua

Kenmore, P (2002). Integrated Pest Management. International Journal of Occupational
& Environmental Health 8(3):]73-174.

MAGFOR (2003). Ministerio de Agricultura, Ganaderia y Forestales. Informe Anual
Agropecuaria 2002-2003. Managua

Qamar, K (2002) Global trends in agricultural extension: challenges facing Asia and the
Paciﬁc region. Food and Agriculture Organization of the United Nations URL:
http://www.fao.Org/sd/2002/kn0903a2fenhtm.

Rivera, W. (2001). The invisible frontier: the current limits of decentralization and
privatization in the developing countries: In E. Brewer (Ed). Agricultural
Extension: An International Perspective. Erudition Press, North Chelmsford, MA.

Sulaiman, V. and A. Hall (2002). Beyond technology dissemination: Reinventing
agricultural extension. Outlook on Agriculture 31(4):225-233.

World Bank (2002). Summary report of the World Bank/USAID/Neuchatel international
workshop on Extension and Rural Development held at IFPRI. Washington DC.

World Bank (2003). Nicaragua poverty assessment. Raising welfare and reducing
vulnerability. Report N° 26128-NI, Washington DC.

DO ENVIRONMENTAL AND HEALTH CONCERNS INFLUENCE FARMERS’

INPUT DECISIONS? EVIDENCE FROM PESTICIDE USE AND ADOPTION OF

INTEGRATED PEST MANAGEMENT IN NICARAGUA

ESSAY ONE

1. Introduction

An assumption underlying many Integrated Pest Management (IPM) diffusion
programs has been that the environmental and health (E&H) effects of pesticide use will
encourage farmers to reduce the use of agrochemicals and to adopt alternative IPM
practices. Although farmers’ awareness of these pesticide-ascribed E&H effects is
extensively recognized, the inﬂuence of this awareness on farmer production decisions
has not been clearly established.

The international development literature has largely discussed the nature of
decisions by households, which naturally produce and consume at the same time. The key
issue has been whether production decisions can be made independently from
consumption and other utility-related decisions. If input and output markets ﬁmction
perfectly, farmers can make production decisions independently of the potential effects
that these decisions could have on household welfare (Singh et a1 1986). But if farmers
face market imperfections or other resource constraints, optimal production decisions
may entail meeting household consumption objectives without market intermediation
(DeJanvry et a1 1991, Strauss and Thomas 1998). The absent markets for ecosystem
services and imperfections in the markets for preventative and curative health care make
these attributes likely to be nonseparable from proﬁtability-related attributes in farmers’
agrochemical input demand decisions.

Evidence of the relationship between E & H effects, and farmers’ production
decisions is scarce and contradictory. While Pitt and Rosenzweig (1986) found that farm
proﬁts among Indonesian rice growers were independent of household health changes

that affect their utility, Van Dusen and Taylor (2005) provided evidence that crop

diversity on Mexican farms is not only a consequence of production characteristics, it
also depends on migration, wealth and farmers’ age and education. Also Beach and
Carlson (1993) found in a hedonic analysis of herbicides that there is a statistically
signiﬁcant relationship between farmer herbicide selection and the user safety and water
quality characteristics that herbicide inputs embody; however both turned out to be minor
components in this production decision. In an IPM diffusion setting, although negative
health effects of pesticide use can affect farm productivity (Antle & Pingali 1994) there is
no evidence on how pesticide-ascribed E & H effects affect farmers’ decisions about
pesticide use, alternative pest control adoption and other production decisions.

This paper uses farm level data from Nicaraguan bean growers in order to test
econometrically a) whether Nicaraguan farmers are aware of the E & H effects of
pesticide use, b) whether farm households have experienced these effects and c) whether
awareness and/or experience of E & H effects inﬂuence farmers’ decisions on pesticide
use, adoption of IPM practices, and family labor allocation. The unexpected results have
important repercussions for ongoing IPM training programs across Central America and
perhaps more widely.

2. Pesticide use, environmental and health effects, and the adoption of Integrated
Pest Management in Nicaragua

Central American agriculture is characterized by high pesticide use compared to
most developing countries. In Nicaragua, a recent survey reported that 88% of small
farmers use pesticides while only 8% use non-chemical pest control (4% did not report
any pest control)(MARENA 1999). Pesticide imports in the region grew from 34 to 45

million Kg between 1994 and 2000. This situation makes Central America the area with

the highest rate of pesticide consumption per capita in the developing world with 1.5 Kg
per person per year (PAHO 2002).

The relatively high pesticide consumption among small farmers has started to
harm human health and insect biodiversity in this area. The Pan American Health
Organization (PAHO) estimates that incidence of acute pesticide poisoning (APP) in
Central America is roughly 20 cases per 100,000 population. In Nicaragua and El
Salvador this rate is estimated to exceed 35 cases per 100,000 population (PAHO 2002).
Worse yet, a recent study estimates that underreporting in the region’s official statistics
approaches 98%, implying that 400,000 poisonings may occur each year with 5% of
people exposed to pesticides experiencing illness symptoms (PANNA 2002). It is also
reported that farmers in Central America recognize that the overuse of pesticides is
destroying the beneﬁcial insect population (Bentley and Andrews 1996). These facts have
encouraged the search for alternative less harmful pest management options.

Integrated Pest Management (IPM) is a group of pest control methods aimed to
reduce H & E risks by keeping a crop pest infestation below an economic threshold level:
the pest level at which control measures are necessary to prevent decline in net returns
(F ernandez-Comejo et a1 1998, Bajwa and Kogan 2002). Pest control methods may
include pesticides when necessary but also non-chemical inputs and specialized practices
such as agro-ecological analysisl, botanical insecticides, insect sticky traps and others
(F emandez-Comejo et al 1998). The combination of proﬁtability and lower E & H risks

of IPM is expected to encourage small farmers to adopt broadly IPM practices.

 

‘ This practice consists in a permanent observation of the crop ﬁeld and the use of insect scouting as the
main activity. It will determine whether pest infestation has reached the economic threshold level.

10

IPM training has a history of more than 20 years in Nicaragua and the rest of
Central America. Many research and development institutions have been developing
extension programs targeted to reduce the use of pesticides and to disseminate IPM
(Staver & Guharay 2003). However, pesticides remain the preferred pest control, and the
adoption of IPM practices has been low in Nicaragua (PROMIPAC 2001, Bentley et a1
2001). In recent years the Program for IPM in Central America (PROMIPAC) has
proposed Farmer Field Schools (FF S) as an alternative extension method that could
produce a broader IPM adoption.

FF S were ﬁrst established in Asia at the end of the 1980’s with the intent to
increase the delivery of IPM by using participatory techniques and the “learning by
doing” approach (Gallagher 1998). The ﬁrst attempts to measure FF S impacts reported
positive results that led development agencies to support FF S implementation worldwide
as the preferred method for delivering IPM (Feder et al 2004). This new extension
approach started in Central America in 2001, but whether and why it is or is not effective
remains to be determined.

3. The analytical framework
3.1 Farmers’ awareness and experience of E & H effects

The ﬁrst step toward identifying the correlation between pesticide-ascribed E & H
effects and farmers’ production decision is to determine the extent of farmers’ awareness
and experience with these E & H.

The literature about pesticide use ﬁnds most farmers to be aware of the potential
E & H risks from pesticide use (Van der Hoek et a1 1998, Heong et al 2002) with some

exceptions in the Aﬁ'ican continent (Maumbe & Swinton 2003). This awareness can

11

result from farmers’ exposure to training or informative programs, learning from
neighbors’ experiences or from direct experience of these effects. However, each of these
can produce different perceptions of the E & H effects and hence different farmer
reactions. Tucker and Napier (2001) found that although some Midwestern US farmers
were aware of potential negative consequences of pesticide use, they doubted the
seriousness of the E & H hazard and still relied heavily on chemical control. Other
studies have also reported a general awareness of pesticide secondary effects, but it seems
that this awareness does not necessarily inﬂuence farmers’ production behavior
(Ecobichon 2001, Heong et a] 2002).

Given the importance of awareness and perception of pesticide E & H hazards for
management decisions, this article will explore the level of Nicaraguan farmers’
awareness. It will also examine the effect of having directly experienced acute E & H
hazards of pesticide exposure. But in order to determine whether pesticide E & H effects
inﬂuence farmers input decisions, a conceptual model of farmer behavior is needed.

3.2. Chemical and non-chemical input decisions among small farmers

A household model provides a framework to understand how small farmers make their
decisions about agrochemical use, the adoption of alternative pest control options and
labor allocation. The household production model approach (Singh et a1 1986) can be
used to analyze the interaction between pesticide use and alternative pest control adoption
(typical production decisions) and the pesticide-ascribed health and environmental effects

experienced by household members. Consider the household utility function as:

U=U(X,,X2,€,H) (1)

12

Where the household derives utility from consumption of on-farm and off-farm
goods (X1 and X2 respectively), leisure (8) and household health (H). Let on-farm goods
and household health can be produced according to the following household production

functions:

QXI =QX1(L(H)’ZC’ch’E’A / CH) (2)

where E=E(E0,Zc / CH)

H=H(X,,X2,Zh,Zc / CH) (3)

The production of on-farm goods (Qx1) depends on labor (L) that could be family
(Lf) or hired (Lh), chemical inputs (Zc), non chemical inputs (ch), environmental services
(E) and ﬁxed inputs (A) like land. Environmental services do not directly affect
household utility, but they beneﬁt household welfare by enhancing crop production
(an). E is represented by beneﬁcial insects which have a natural endowment (E0) and are
affected negatively by chemical use (E’(Zc) S 0). All of these inputs contribute positively
to crop production. Health is expected to contribute to farm labor supply (L’(H)>0).
Health can be augmented by consuming goods (especially food) and health inputs (2,.)
like health care services and protective devices, but health is diminished by use of
chemical inputs (H’(Zc) S 0). The three production functions are inﬂuenced by exogenous
household characteristics (CH), socioeconomic, cultural and otherwise.

The household faces an income constraint represented by:

er-XI+ Px2.x2 + PZc-Zc+ Pch-ch+ PZh-Zh S H + Y (4)

l3

Where Pm and Ba, are the output market prices, PZC, Pch and P2}, are input
market prices, Y is non-fann income, and gross revenues are deﬁned as: II = PX]. X1 -
w*L. Thus gross revenues are conditional on shadow wage (w*) which is the relevant
measure of the marginal cost of labor under labor market imperfections (Singh et a1
1986) like those in Nicaragua.

The optimal solution to this constrained utility maximization problem depends on
whether farmers make input use decisions considering their potential impact on
household welfare (like on household health and environmental services) or whether
these decisions are made independently meaning that production and consumption
decisions are separable. This paper will focus on the chemical input use, IPM adoption
(non-chemical inputs) and labor allocation optimal decisions.

If production and consumption decisions are made jointly (non-separably) and
farmers believe that pesticide use can produce secondary effects, the optimality

conditions for chemical use, non chemical use and labor use imply the following:

6U 6H 6U 6E +6U 6L 6H

 

 

 

 

 

+
59 6H 62, 6E 62 6L 6H 62
P = P2 — c c
" azc ‘ A. (5)
.61: 69
aQ _ _ 69 62,,
Pg 62 — Pch A. (6)

14

 

 

 

L,

2‘1. 1Q.
1,1533%: w _ anaL (7)

Now the marginal value product of chemical inputs (Zc) no longer equals just its
market price, but rather its market price mg the marginal effects of pesticide use on
household utility through E & H, divided by the marginal utility of income (A). If farmers
perceive the negative impact of pesticide use on health and natural environment, they will
reduce pesticide use from the proﬁt maximizing level. In the case of non-chemical inputs
and labor allocation optimal conditions, the marginal cost in both cases also includes the
marginal effect of both production inputs on household utility. Of course, if production
and consumption decisions are separable (Singh et al 1986), and farmers believe that E &
H effects will not affect their household welfare, then the marginal value product of each
production input would equate, as usual, to its market price.

Evidence exists both for and against separability of pesticide input decisions from
household welfare. Many studies have referred to developing countries as regions where
markets are underdeveloped and subject to endemic imperfections like inadequate access
for small farmers, lack of information, high transaction costs or input heterogeneity
(DeJanvry et al 1991 , Van Dusen and Taylor 2005). The non- separability hypothesis
under these conditions is likely to hold. On the other hand, many development projects
target market-integrated farmers who have permanent access to product markets, on-farm
and off-farm work opportunities and reasonable access to technical assistance, health
services and credit markets. Also recent studies have shown that farmers are able to value
some of the E & H effects related to pesticide use (Higley and Wintersteen 1992, Cuyno

et al 2001). These conditions can support the competing separability hypothesis.

15

A better market integration could lead small farmers to use available averting
measures to avoid E & H effects of pesticide exposure and make them believe that these
effects would not really affect their household welfare. Households could hire labor for
pesticide application, avoiding the negative effects on their family members. They can
also have access to protective devices (i.e. masks, gloves) or to curative health care if
farmers get poisoned reducing the potential effects of pesticides on household health.
Finally, if farmers are able to value ecosystem services and health effects, these farmers
will be able to compare their valuation of the hidden cost of E & H effects with the
potential economic loses if they modify their proﬁt maximizing input decision. Whether
farmers consider E & H effects on their input decisions become an empirical question and
we develop a separability test in order to evaluate the correlation between the pesticide-
ascribed E & H effects and the household demand for pesticides, IPM practices and labor.
4. Model empirical implementation
4.1. The separability test

The separability hypothesis can be tested by examining whether the marginal
value product of each input equates to its market price (chemical input price, non
chemical input price and wage) (Benjamin 1992). The alternative hypothesis in all the
cases is that farmers deviate from the standard rule, and this deviation is correlated with
variables that affect household welfare (E & H effects and household characteristics). In
other words, the alternative hypothesis implies that each input’s marginal value product
equates to an input shadow price that depends on variables that affects household welfare.
Deﬁning B and y as the vectors of coefﬁcients that will be estimated for health effects and

environmental effects, the null hypothesis can be stated as:

16

H0=B=y=0

If we fail to reject the null hypothesis, there would be evidence that market
imperfections cause farmers to consider health and environmental effects while making
decisions about pesticides, adoption of IPM practices and the use of labor for spraying
chemicals.

For implementing this test we derive the following input demand functions for

pesticides, IPM practices and labor for applying pesticides from the optimal conditions

described above:
2, =ZC(RI,PZC,PZnC,w,A,H, E; C”) , (8)
_. . (9)
ch _ch(}2]’PZc’Pch’ W’A’H’ E’ CH)
L _.L . (10)
d — d(l:,ch,Pch,IV,/I,H,E,CH)
4.2. The Data

Farm level data for this study were collected between May and August 2004 via a
cross-sectional survey of 436 households of Nicaraguan bean growers in the departments
of Esteli, Matagalpa, J inotega, Madriz, Masaya and Carazo. The survey was designed
following a double stratiﬁcation (Deaton, 1997) based on exposure to IPM training
participation in FFS and other programs. Households in 74 rural communities were
interviewed, including 13 where FFS were implemented, 9 where FFS graduates lived but

no F FS were held, 26 communities selected randomly where no FFS exists but where

17

IPM extension services were available, and 26 communities selected randomly where no
IPM extension was present. In each community, households were selected randomly and
included both clients and non clients of NGOS. The sample distribution includes FF S
graduates, farmers participating in other IPM programs, FF S graduates who also attended
other IPM programs, and farmers who had no prior contact with formal IPM extension.
4.3. The econometric estimation

4.3.1. Potential econometric problems

The sample design poses secondary econometric problems of a clustered and
stratiﬁed sample that can bias the parameter estimates (Wooldridge 2002a). To deal with
the clustering problem, this paper adjusts the variance matrix and includes dummy
variables for each cluster using survey regression methods (Wooldridge 2003). This
method is suitable given the nature of the sample design that has small clusters (six
households per village on average) and a large number of clusters (74 villages). To
correct the unbalanced representation of farmers (especially FF S trainees) in the sample,
this paper uses a weighting scheme as suggested by Wooldridge (2002a) (see appendix
A3 for details).

Another potential problem for the econometric estimation is the possibility of
measurement error of the E & H variables that can bias the estimation of their effects
over the production decisions, and therefore invalidate the conclusions about separability.
However as measurement error is correlated with the true E & H variables and not with
the observed variables in the sample, standard models like OLS and multinomial logit can

still be produce consistent estimates (Wooldridge 2002) see appendix A.4 for details).

18

4.3.2. The econometric specification
This paper speciﬁes four linear models for pesticide demand, a multinomial logit
model for the adoption of IPM practices, and three linear models for the labor use. All

models have the same set of explanatory variables and the general model structure is:

X]. =13d6jp+H,,,Bj,7,+E,6]E+C,,,BJ.CH+CC,BJ.CC+Uj ;forj=1,...,8 (11)

The J dependent variable depends on vectors of k output and input prices (Pk), h self-
reported past acute health symptoms (Hh), perceived beneﬁcial insect population levels
(E), socioeconomic and other household characteristics (CH) and community ﬁxed effects
(Cc), with disturbances assumed to be independently distributed (U i).All variables used in
the econometric estimation are described in detail in the following subsections.

4.3.3. The dependent variables

Household demand for insecticides, herbicides, fungicides and molluscicides2 are
represented by the quantity of pesticide active ingredients used by each household in
bean production during the last season weighted by acute human toxicity. A human
toxicity index is calculated for each household i as the sum over all m pesticide active
ingredients (aiik in kg/ha, k=l ,. . .,m) divided by each active ingredient’s mammalian
toxicity (measured by the minimum dose per gram of body weight that is lethal to 50%
of a test rat population or LDSO, as reported in USDA, 1998). The human toxicity index,

show below, is proportional to the LD50 and is increasing in lethality.

 

2 Insecticides are mainly used to control white ﬂy (Bemisia tabaci (Genn)) and diabrotica beetles
(Diabrotica Sp); fungicides to control anthracnose and angular leaf spot, and molluscicides to control
mainly slugs (Vaginulus plebeius (F isher)). The main active ingredients used in Nicaragua are
metamidophos (insecticides), paraquat (herbicides), mancozeb (fungicides) and metaldehyde
(molluscicides).

l9

 

at.
Mimi,

The adoption of IPM practices is represented by binary variables that indicate
whether a household adopted individually or simultaneously agro-ecological analysis,
botanical insecticides and/or yellow insect traps. These are the most promoted IPM
practices in Nicaragua and the only three that were common across different IPM training

programs and within each of them. The latent variable is deﬁned as:

chj = 0 if household did not adopt any IPM practice
1 if household adopted only insect scouting
2 if household adopted only botanical insecticides
3 if household adopted only yellow traps
4 if household adopted a combination of two practices

5 if household adopted the three IPM activities

Finally, labor demand is represented by the natural logarithm of the family labor,
hired labor and total farm labor.
4.3.4. The explanatory variables
Output prices are selling prices for bean and maize reported in USS/kg. Input prices
include bean seed, metamidophos insecticide, gramoxone (paraquat) herbicide and

mancozeb fungicide3. The bean the net revenue function also included the price of

 

3 Metamidophos, paraquat and mancozeb were by far the main active ingredients among insecticides,
herbicides and fungicides on the Nicaraguan farms surveyed in 2004

20

fertilizers‘. Other prices and price proxies are the transport cost to the nearest market of a
sack of beans or maize and the distance between each farm and the municipal capital.
Other production variables include the most recent farming season that bean were grown:
primera, postrera or apante, the use of disease resistant bean varieties, altitude (proxy for
rainfall and temperature) and farmers’ observed level of bean pests and diseases (high or
not).. For pesticide demand, we also include the adoption of the three deﬁned IPM
activities that were observed to occur prior to pesticide use decision. We also include a
variable for exposure to IPM extension training.

Health variables in this paper refer mainly to past health events or pre-existing
conditions at the beginning of the most recent bean season prior to the 2004 survey. As
97% of the surveyed households reported awareness of pesticide-ascribed health
problems, this variable is dropped from the econometric analysis because it could not
explain pesticide use or IPM adoption variability. We do include the number of acute
health symptoms suffered by household members in the past and the frequency of the
following acute symptoms: diarrhea, respiratory difﬁculties, head ache, dizziness,
vomiting, stomach ache, blurred vision, eye irritation and skin rash. Symptom
frequencies measure four level of incidence of a speciﬁc symptom after applying
chemical inputs: never, rarely, sometimes or frequently. Other health variables include
whether household members had to visit a city doctor or become hospitalized due to
pesticide poisoning, whether these household members lost workdays after pesticide
poisoning, whether the household hired labor or used family labor for most pesticide

applications, and how many purchased or home-made protective devices (e. g. mask,

 

‘ we use the price of the most common combination of nitrogen, phosphorus and potassium (12-30-10)

21

gloves or special clothing) for applying chemicals were on hand before the most recent
bean production season.

The environmental variables are represented by the respondent’s awareness of
beneﬁcial insects and the reported level of beneﬁcial insect population in the previous
bean season. Although many farmers could measure whether this population was greater,
lower or unchanged compared to the previous season, almost all of the respondents to this
question were farmers who had received IPM training. For this reason we include these
variables as an interaction with IPM training participation and with IPM adoption (Table
1.1). For the same reason beneﬁcial insects are not included in the IPM adoption model.

Household characteristics are represented by two groups of variables: (1)
measures of household assets: farm size, area under irrigation, number of cattle, whether
family has electricity at home, and whether the household receives remittances from
relatives working in foreign countries; and (2) measures of household composition:
gender of household head, age and years of education of household head, percentage of
females, and percentage of family members under 14 years old. Finally dummy variables
are included for community-level ﬁxed effects with one of them kept as a control. The 73
community dummies are used for the pesticide and labor demand, but are substituted by
municipalin dummy variables (9 out of the 10 municipalities) in order to avoid the
computational problems of the multinomial logit estimation in the IPM demand.

5. Results
5.1. Farmer awareness of pesticide-ascribed health and environmental effects
Most of the respondent households used pesticides during the previous bean

season. Seventy ﬁve percent applied insecticides, 60% herbicides, 22% fungicides and

22

18% molluscicides. This pesticide use was accompanied by uneven farmers’ awareness
of potential health and environmental effects. While 97% of the interviewed households
were aware of the potential negative effects that pesticide use can have on human health,
only 38% recognized the existence of beneﬁcial insects and the possibility that these
beneﬁcial insects can help to control farm pests (Table 1.1). Large educational campaigns
carried out by the Nicaraguan government and some international and national
development agencies explain the high farmer awareness of these health effects
(PLAGSALUD 1997), but information about beneﬁcial insects has only been diffused by
limited IPM training efforts in Nicaragua (Bentley and Andrews 1996). If we consider the
over sampled number of households exposed to IPM training (Table 1.1), the true level of
awareness of beneﬁcial insects would be much lower.

The prevalent awareness of potential pesticide risks to human health among
Nicaraguan farmers rules out lack of awareness as a plausible explanation for not
adopting pesticide risk reduction strategies, including IPM practices. For this reason this
variable will be dropped from the econometric analysis.

5.2. Farmers’ direct experience with environmental and health effects

At least 68% of the farm households reported having suffered at least one of a
variety of acute illness symptoms and to have experienced on average 3 symptoms after
applying pesticides. The most common symptom reported was headache (48% of
respondents) and the least common was diarrhea (only 3% of respondents). The
frequency of these symptoms was far from homogenous. While most of the symptoms

were never experienced by households, headache and eye irritation were reported to be

23

experienced “always” by 19% and 16% respectively of the households after applying
pesticides.

Table 1.1. Summary statistics on key variables. 436 Nicaraguan bean growers, 2003-
04+

 

 

Variables Mean Std Dev.
Dependent Variables
Index of active ingredients of insecticides weighted by toxicity (x 100) 2.9 5.1
Index of active ingredients of herbicides weighted by toxicity (x 100) 0.5 0.7
Index of active ingredients of fungicides weighted by toxicity (x 100) 0.1 0.6
Index of active ingredients of molluscicides weighted by toxicity (x 100) 0.1 0.2
Adopted only agro-ecological analysis (%) 4.4
Adopted only botanical insecticides (%) 1 1.7
Adopted only insect yellow traps (%) 2.8
Adopted two IPM practices (%) 14.9
Adopted three IPM practices (%) 10.8
Quantity of family labor for applying agrochemicals (man-days) 3.1 3.4
Quantity of hired labor for applying agrochemicals (man-days) 0.8 1.6
Health variables+
Awareness of pesticide ascribed health effects (%) 96.8
Households who suffered from

Headache 48.2

Eye irritation 43.1

Stomachache 23.4

Skin rash 29.0

Blurred vision 27.6

Muscle pain 26.2

Respiratory difﬁculties 1 1.2

Diarrhea 3.2
Visited city doctor after pesticide poisoning (%) 6.8
Hospitalized aﬁer pesticide poisoning (%) 5.3
Reported workdays lost after pesticide poisoning (%) 17.4
Number of purchased protective devices 0.4 0.7
Number of homemade protective devices 0.3 0.5
Households hire a pesticide applicator (%) 18.0
Beneﬁcial insect variables
Awareness of beneﬁcial insects (%) 38.0
Awareness of beneﬁcial insects and IPM training 33.0
Households that observed a higher level of beneﬁcial insects (%) 8.0
Observed higher beneﬁcial insects and adopted AAE 8.0

 

+ Details about other variables used in the econometric analysis can be found in detail in
Appendices A5 and A6

24

The past experience of interviewed households also includes severe cases of acute
health illness aﬁer applying pesticides. Seven percent of the farmers had to go to a city
doctor after getting poisoned and 5% had to be hospitalized for the same reason (Table
1.1). Getting in to a hospital could take up to more than 4 hours for some of the places we
visited. In addition, 17% of those interviewed reported having lost work days due to
pesticide poisoning.

Of the 62% of the sample that received IPM training and the 38% that recognized
the existence of beneﬁcial insects, only 22% were able to observe beneﬁcial insects
during the previous bean season and to evaluate whether the level of these beneﬁcial
insect had changed. Of that 22%, 8% of the households observed an increase in the level
of beneﬁcial insects, 10% the same level and 4% a lower level compared with the
previous season.

5.3. The influence of E & H effects on household production decisions
5.3.1. Demand for pesticides

Pesticide ascribed E & H effects inﬂuence use of the four classes of pesticides in
different ways. Only in the instance of insecticides could we reject the null hypothesis
that E & H effects jointly have no effect on the household demand for pesticides (Table
1.2). We would reject this null hypothesis in the herbicide case with only 82% conﬁdence
and cannot reject the null hypothesis of no E & H effect in the instance of fungicide and
molluscicide use. These results are consistent with the fact that the most toxic pesticides
used in Nicaragua are insecticides (metamidophos) and herbicides (paraquat). Table 1.3
summarizes the regression results of the four chemical demand models which display

good ﬁt considering the cross sectional nature of the data.

25

Table 1.2. The jointly effect of E & H variables on household input decisions, F-test
for 436 Nicaraﬂan bean growers, 2003-04

 

 

 

Model F-statistic P-value
Pesticide demand
Insecticides 1 .91 0.03 17
Herbicides 1 .37 0.1 852
Fungicides 1 .06 0.3970
Molluscicides 0.95 0.5132
IPM adoption 65.76 0.0000
Labor demand
Family 3.87 0.0000
Hired 1.87 0.0295
Total 2.10 0.0123

 

The toxicity-weighted demand for insecticides was signiﬁcantly affected by prior
experience with individual acute poisoning symptoms and farmer awareness of beneﬁcial
insects (Table 1.3). However the magnitude of these effects measured as standard
deviation (s.d.) of the dependent variables are relatively small (Table 1.3). Households
that had suffered more frequent diarrhea and skin rash reduced the toxicity weighted use
of insecticides by 0.08 and 0.02 s.d., but households that suffered more frequent
headaches signiﬁcantly increased the use of these insecticides by 0.03 s.d. This
contradictory behavior is explained by farmers’ explanation that headaches are difficult
to avoid after insecticide use (Table 1.1), but they are easily treated with headache pills.

Households that had more severe health effects and experienced workdays lost
after applying pesticides also reduced the use of insecticides during the previous bean
season by 0.05 s.d.. Farmers who were aware of beneﬁcial insects reduced their use of

insecticides by 0.07 s.d.. The insecticide reduction response came only among

26

respondents aware of beneﬁcials who had IPM training, not the subset who had adopted

agro-ecological analysis.

Table 1.3. Inﬂuence of E & H effects on pesticide use, Linear survey regression
results for 436 Nicaraguan bean growers, 2003-04+

 

 

 

Insecticide Herbicide Fungicide Molluscicide
IPM trained & beneﬁcial -0.336* -0.059
awareness
(2.96) (0.78)
AAE & more beneﬁcials -0.171 0.337
(1.16) (1.32)
Number of symptoms 0.012 0.070 -0.003 -0.010
(0.37) (1.37) (0.14) (0.39)
Frequency of eye irritation -0.009 -0.027 0.002 0.002
(0.15) (0.47) (0.04) (0.04)
Frequency of diarrhea -0.398* -0.048 0.013 -0.118
(1.74) (0.47) (0.13) (1.01)
Frequency of respiratory difﬁculty 0.084 0.012 -0.020 -0.055
(0.86) (0.14) (0.35) (0.62)
Frequency of skin rash -0.106* -0.028 ~0.017 0.013
(1.65) (0.52) (0.42) (0.28)
Frequency of stomach ache 0.039 0.035 -0.067 -0.048
(0.54) (0.33) (1.58) (0.62)
Frequency of head ache 0.177" 0.002 -0.050 -0.029
(2.26) (0.03) (1.41) (0.60)
Frequency of muscle pain 0.052 -0.012 0.088 -0.040
(0.87) (0.21) (1.47) (0.74)
Frequency of blurred vision -0.017 -0.137* -0.029 0.060
(0.16) (1.98) (0.56) (0.61)
Lost workdays -0.233* -0.033 -0.000 -0.040
(1.65) (0.30) (0.00) (0.43)
Visited city doctor -0.022 0.078 0.051 -0.031
(0.13) (0.54) (0.39) (0.26)
Hospitalized -0. 128 0.241 0.068 -0.053
(0.52) (1.53) (0.40) (0.39)
Purchased protective devices 0.212*** -0.046 0.028 -0.003
(3.57) (0.85) (0.78) (0.08)
Homemade protective devices -0.l78* -0.146 0.039 -0.034
(1.73) (1.49) (0.45) (0.41)
Hired applicator 0.189* 0.222* 0.082 0.104
(1.89) (1.72) (1.06) (1.08)
R-squared 0.46 0.52 0.37 0.39

 

* Signiﬁcant at 10%; " signiﬁcant at 5%; *" signiﬁcant at 1%
Absolute value of t statistics in parentheses

+ Complete regression results are Shown in apmndix A.7

Although the evidence shows that Nicaraguan bean growers reduce insecticide use in

response to direct experience with acute illness symptoms and beneﬁcial insects,

parallel evidence shows that when averting measures to avoid pesticide exposure are
available in the market, households will buy them and keep using more pesticides.
Farmers who use more purchased protective devices or hired a pesticide applicator
increased their use of toxicity-weighted insecticides by 0.04 s.d. in both cases. On the
other hand, farmers who used homemade protective devices reduced their use of
insecticides (Table 1.3). The opposite effects of protective devices may be explained by
the lower quality of homemade products like simple plastic or handkerchiefs to cover the
body or face, which may not reduce the toxicity risk of pesticide exposure.

Although E & H effects do not jointly inﬂuence the herbicide use decision,
households that had more frequent suffered blurred vision decreased their use of
herbicides by 0.03 s.d. Households that hired a pesticide applicator also increased the use
of herbicides by 0.04 s.d.

5.3.2. IPM adoption

The adoption of IPM practices is also signiﬁcantly inﬂuenced by prior experience
with pesticide poisoning symptoms. We reject the null hypothesis that all health variables
jointly have no effect on the adoption of agro-ecological analysis, botanical insecticides
and insect yellow traps (Table 1.2). Table 4 shows the results of the multinomial logit
estimation of this adoption decision. As the survey multinomial logit results could not be
expressed in marginal probabilities, we only discuss the Sign of these results.

Although only 4% of respondents adopted only agro-ecological analysis (Table
1.1), various acute illness effects and the severity of these symptoms inﬂuence
signiﬁcantly and positively the adoption of this IPM practice. An extra number of acute

symptoms and having experienced more frequently skin rash make households more

28

likely to adopt agro-ecological analysis (Table 1.4). Households that faced severe cases
of pesticide poisoning and had to visit a city doctor or be hospitalized were also more
likely to adopt this IPM practice. The fact that some households lost work days due to
pesticide poisoning seems to reduce the adoption of agro-ecological analysis. Finally, a
greater use of purchased protective devices seems to complement IPM adoption, because
a farmer who increased the use of these devices increased the probability of IPM
adoption.

The use of botanical insecticides is the IPM practice with the greatest individual
adoption (11.7%), which is consistent with the history of IPM training in Nicaragua
promoting this practice. However, the likelihood of adopting this practice is affected
differently by experience with different acute symptoms. While an extra number of acute
symptoms or speciﬁcally having suffered more frequently headache made households
more likely to adopt botanical insecticides, households with more frequency of eye
irritation, dizziness and blurred vision were less likely to adopt these botanical
insecticides (Table 1.4). This partial discouragement of adopting botanical insecticides is
associated with some negative effects that the use of some natural products can have over
human health. Many botanical insecticides are made base on chile (Capsium anum),
manure, alcohol and some plants like Gliricidia cepium or Azadarichata indica that can
affect the applicator’s health.

Yellow sticky traps for insects constitute the least adopted individual IPM
practice (2.8%) and it is not inﬂuenced by past acute illness symptoms experienced by

households (Table 1.4). Most of the 16% of respondents who reported having adopted

29

 

.u.< 585%? E 5505 0.3 3:52 ecumbE 30.8800 +

momoﬁﬁoaa Gm moﬂmmuﬂum a MO 029; Qua—Omn<

so. a “585:3... .: can 3 assumes: : was 3 2885.5 .

 

30

 

~58 628 pm. 8 8.8.8 58
8a.? 5.8 $3“- :52 ES. BEBE 8::
A38 8°. 8 2 n8 A28 $8
:33 :2 «an: :2 $3 8038 283E ”Basso:
:38 8N8 an. 8 3.8 as. 8
3.3 33 5:: 83 has... 883% Essen 8855
$8.88 38.8 8&8 398 $28
2:32.. .23. 88..» $3 32.2... sagas:
:98 5.. 8 Go. 8 3.8 as. 8
583- £3. 83- 82- .52». 38% 58 33>
3&8 A38 am. 8 $8 as. 8
82. 23. a _ 02. 3:8. :23- $353 23
808 an 8 E8 £88 an. 8
:5..- 3.8 82.8- :§._- 83- 8s; 853 cc 55:8;
5.8 $08 8.8 5.8 3.4.8
83. :38 «2.8 $2.. 32 Ea 28:5 co 55:3;
£8 5. 8 Go. 8 5.8 8.98
:2»... .883 $3 :38... 83. £8 82 co 55:3;
A38 §8 AS. 8 38 2:8
8 _ a 82. 88.8 S .3. $2. 2.8 55% co 55:3;
:38 A38 8.. 8 8.8.8 5.8
5.5...— SS 9%.: 3:. .3»... 5.2 83° 5:33....
$8 and am. 8 5.8 68.8
$2 Ema.— ..S.» _- and 5:. 3.88:. reﬁnance 55:3:—
am. 8 an. 8 §. 8 5.8 :68
$2. .33.. 88.2 5:84- $28. 30553.. @532“.
3.. 8 $8 2:. 8 5.8 $8
Ste ”2.? 43.9 :53. $2 SEE ”a co 3:83;
888 5.8 as. 8 308 E8
238 $3 RE :33... 5.3m... seamen co .352
acute-En 2.: 09:; 39:99:. E& 95,—. was... 38:— 823385 .3382— m_m>_a=< ._oue.c..u<

 

 

+34”ch .9532 53.— :a: «.332
one .8 £32 a»... 1328...... 8:5 .8568 2.: 2.935.. as 5:83 2.. .5 uses a a m as 85.5.: :93;

insect traps did so jointly with agro-ecological analysis or botanical insecticides. The
simultaneous adoption of two IPM practices is also affected by past acute health
symptoms but differently. While greater frequency of respiratory difﬁculties increased
the probability of adopting IPM practices, greater frequency of dizziness reduced this
probability. This last effect was associated with the main adoption of botanical
insecticides, which as noted could produce acute health symptoms.

Finally, households having experienced a greater frequency of skin rash,
headache, and hospitalization after getting poisoned were more likely to adopt the three
IPM practices (Table 1.4). As the adoption of botanical insecticides is reduced among
households having suffered of blurred vision, the same is true of the adoption of all three
IPM practices (Table 1.4). Similarly, households having visited a city doctor due to
pesticide poisoning are less likely to adopt the three IPM practices that include botanical
insecticides presumably due to the latter’s potential negative health effects.

5.3.3. Labor demand

E & H effects also significantly inﬂuence the demand for family, hired and total
labor for applying pesticides. All three labor demand models produced R squared values
over 0.35, implying sound explanatory power (Table 1.5).

Households that had suffered from blurred vision reduced by around 0.6 the
number of family labor days per hectare spent applying chemicals (0.18 s.d. of this
dependent variable). They also reduced by 0.7 man-days per hectare the hired labor for
applying pesticides (0.43 s.d.). By contrast, households that had suffered a greater
frequency of eye irritation increased the use of hired labor for applying pesticides by 0.6

man-days per hectare (0.4 s.d.). The frequency of prior diarrhea linked to pesticide use

31

had no effect on either family or hired labor, but it did reduce total labor on beans by 0.9

man-days per hectare (0.18 s.d.) (Table 1.5). Behind this aggregate effect, there were

many households that responded to having suffered diarrhea by reducing the use of

family labor and increasing the use of hired labor.

Table 1.5. Inﬂuence of E & H effects on labor demand for pesticide application.
Linear survey regession results for 436 Nicaraguan bean growers, 2003-04+

 

 

 

Family labor Hired labor Total labor
IPM trained & beneﬁcial awareness -1.856*** 0.778 -0.894
(2.87) (1.12) (1.63)
AAE adoption & more beneﬁcials -1.872** -0.269 4.636"
(2.36) (0.20) (2.12)
Number of symptoms 0078 0.253 0.015
(0.59) (1.32) (0.15)
Frequency of eye irritation -0.27 7 0.640" -0. 180
(1.20) (2.02) (0.96)
Frequency of diarrhea -0.431 1.032 -0.886*
(0.58) (1.33) (1.67)
Frequency of respiratory difﬁculty 0.095 0.548 -0. 163
(0.28) (1.22) (0.52)
Frequency of skin rash 0169 -0.689* -0.371
(0.64) (1.86) (1.52)
Frequency of stomach ache 0.158 -0.496 0.1 13
(0.54) (1.1 1) (0.45)
Frequency of head ache 0.163 -0.537 0.046
(0.77) (1.60) (0.26)
Frequency of muscle pain 0.346 -0.163 0.245
(1.54) (0.41) (1.33)
Frequency of blurred vision -0.620*** -0.248 -0.259
(2.67) (0.48) (1 .46)
Lost workdays -0.542 -0.766 -0.543
(0.80) (1 .06) (1.00)
Visited city doctor -0.981 -0.144 -0.813
(1.02) (0.16) (0.97)
Hospitalized 0.5 15 -l . 149 0.224
(0.65) (0.84) (0.32)
Purchased protective devices 0.426* -0.287 0.099
(1.70) (1.02) (0.43)
Homemade protective devices -0.712 -0.3 16 -0.765*
(1.56) (0.61) (1.81)
R-squared 0.43 0.37 0.46

 

"‘ Signiﬁcant at 10%; " signiﬁcant at 5%; "* signiﬁcant at 1%
Absolute value of t statistics in parentheses
+ Complete regression results are shown in appendix A.7.

 

32

Awareness of beneﬁcial insects affected only the use of family labor (Table 1.5).
Farmers trained in IPM and who reported being aware of the existence of beneﬁcial
insects reduced the use of family labor in pesticide application by 1.9 man-days per
hectare (0.55 s.d.). The same occurred among farmers who adopted agro-ecological

analysis and who observed a greater level of beneﬁcial insects (Table 1.5).

6. Conclusions

These results have implications for policy makers designing IPM delivery
programs. The ﬁndings support the underlying assumption of most IPM training efforts
that farmers’ awareness of and direct experience with pesticide-ascribed E & H risks can
reduce the use of pesticides and increase the adoption of non-chemical IPM practices.
However the validity of these non-separable assumptions is limited to certain conditions:
First, perceived E&H hazards are only important enough to inﬂuence farmer practices
when farmers are using toxic or highly toxic pesticides. Below that level Nicaraguan bean
growers ignore the potential E & H risks or treat them as separable from their production
input decisions. Second, when farmers face high health risks, they tend to use protective
strategies for avoiding direct exposure to pesticides. In addition to using recommended
protective devices while applying pesticides, Nicaraguan bean growers shift risk by
hiring pesticide Sprayers and transferring the potential health hazards out of the
household. This ﬁnding is consistent with the separability hypothesis because it treats the
health of hired workers as external to household input decisions.

Non-chemical IPM practices can constitute an alternative or a supplement to an
entirely chemical-based pest management strategy. However, some non-chemical

practices, such as certain botanical insecticides, also produce negative health effects. The

33

risks of these practices should be communicated to potential IPM adopters in order to
encourage them to use protective devices to avoid direct exposure.

IPM training efforts need to address all of these challenges and improve their
educational component. Farmers trained in IPM need also to be aware of off-farm and
long run E & H effects that are not easily perceived. This paper did not incorporate these
indirect and chronic effects but farmers with more comprehensive information about
secondary effects of pesticide use might reduce pesticide use even farther and increase
their adoption of non-chemical IPM practices, rather than shiﬁ risk outside of the

household.

34

References

Antle, J and P. Pingali (1994). Pesticides, productivity and farmer health: A Philippino
case study. American Journal of Agricultural Economics 76(3):418-430.

Bajwa, W and M. Kogan (2002). Compendium of IPM deﬁnitions (CID). What is IPM
and How is it Deﬁned in Worldwide Literature? IPPC Publication N° 998.
Integrated Plant protection Center (IPPC). Oregon State University. Corvallis, OR
97331, USA.

Beach, E and G. Carlson. 1993 A hedonic analysis of herbicides: Do user safety and
water quality matter? American Journal of Agricultural Economics 75: 612-623.

Benjamin, D. (1992). Household composition, labor markets, and labor demand: Testing
for separation in agricultural household models. Econometrica 60 (March): 287-
322.

Bentley, .1 and K. Andrews (1996). Through the roadblocks: IPM and Central American
smallholders. International Institute for Environment and Development.

Gatekeeper series N° 56, London.

Bentley, J ., L. Morales and A. Zamora (2001). La Extension de MIP en Nicaragua.
CATIE, Managua. 42pp.

Cobbe, R. (1998). Capacitacio’n Participativa en el Manejo Integrado de Plagas (MIP):
Una Propuesta para América Latina. F AO, Roma, 46pp.

Cuyno, L., G. Norton and A. Rola (2001). Economic analysis of environmental beneﬁts

of integrated pest management: a Philippine case study. Agricultural Economics
25(2):227-233.

Deaton, A. (1997). The Analysis of Household Surveys. A Microeconometric Approach to
Development Policy. The John Hopkins University Press, Baltimore and London.

De Janvry, A., M. F afchamps and E. Sadoulet (1991). Peasant household behavior with
missing markets: Some paradoxes explained. Economic Journal 101: 1400-1417

Ecobichon, D (2001). Pesticide use in developing countries. Toxicology 160:27-33.
Femandez-Cornejo, J ., S. Jans and M. Smith (1998). Issues in the economics of pesticide

use in agriculture: a review of the empirical evidence. Review of Agricultural
Economics 20(2): 462-488.

35

Heong, K., M Escalada, V. Sengsoulivong and J. Schiller (2001). Insect management
beliefs and practices of rice farmers in Laos. Agriculture, Ecosystems and
Environment 92: 137-145.

Higley, L and W. Wintersteen (1992). A novel approach to environmental risk
assessment of pesticides as a basis for incorporating environmental costs into
economic injury level. American Entomologist 38: 34-3 9.

MARENA (1999). Andlisis del Banco de Datos del circa Protegida de Miraﬂor.
Ministerio de Ambiente y Recursos Naturales de Nicaragua, Managua.

PAHO (2002). Pan-American Health Organization. Epidemiological Bulletin 23 Num 3.

PANNA (2002). Pesticide Action Network North America. Pesticide Poisoning-a Central
American Epidemic. Global Pesticide Campaigner Vol 12 num 3.

Pitt, M and M. Rosenzweig (1986). Agricultural prices, food consumption and health and
productivity of Indonesian farmers. In Singh, 1, L. Squire and J. Strauss, (Eds)
Agricultural Household Models: Extensions, Applications and Policy. Johns
Hopkins University Press, Baltimore.

PLAGSALUD (1997). Conocimientos y uso de plaguicidas en poblacio'n usuaria de los
SILAIS de Leon, Estela, Chinandega, Carazo, Rivas, Granada, Masaya y
Matagalpa. Organizacion Panamericana de la Salud and Ministerio de Salud de
Nicaragua, Managua.

PROMIPAC (Proyecto Manejo Integrado de Plagas en América Central) (2001)
Evaluacio’n de metas del proyecto PROMIPAC en su segundafase I998-2001.
Esteli, Nicaragua.

Singh, 1., L. Squire and J. Strauss, eds (1986). Agricultural Household Models:
Extensions, Applications and Policy. Johns Hopkins University Press, Baltimore.

Staver, C. and F. Guharay (2003). Building Integrated Pest Management practices in
Central America. Experiences from CATIE. In K. Maredia, D. Dkouo and D.
Mota-Sénchez (eds) Integrated Pest Management in the Global Arena, CABI
international, Wallingford, UK.

Strauss, J and D. Thomas (1998). Health, nutrition and economic development. Journal
of Economic Literature 36(2):766-817.

Tucker, M and T. Napier (2001). Determinants of perceived agricultural chemical risk in
three watersheds in the Midwestern United States. Journal of Rural Studies 17:
219-233.

36

Van der Hock, W., F. Konradsen, K. Athukorala and T. Wanigadewa (1998). Pesticide

poisoning: a major health problem in Sri Lanka. Social Science & Medicine 46
(4-5):495-504.

Van Dusen, E and J .E. Taylor (2005). Missing Markets and Crop Diversity: Evidence
from Mexico. Environment and Development Economics 10(4): 513-531.

Wooldridge, J (2003). Cluster-sample methods in applied econometrics. American
Economic Review 93 (May): 133-138.

Wooldridge, J. (2002). Econometric Analysis of Cross Section and Panel Data. MIT
Press, Cambridge, MA.

37

DOES PARTICIPATION IN FARMER FIELD SCHOOLS IMPROVE
GRADUATES’ FARM, HEALTH AND ENVIRONMENTAL OUTCOMES? THE

CASE OF NICARAGUAN BEAN GROWERS

ESSAY TWO

38

1. Introduction

The delivery of integrated pest management (IPM) has challenged the
effectiveness of national extension services in developing countries for not achieving
desired adoption levels (Addo et al 2001, Chaves & Riley 2001, Or 2003). IPM is a
complex knowledge-intensive technology that is expected to reduce environmental and
health (E&H) risks associated with pesticide use without affecting crop proﬁtability
(F emandez-Comejo et al 1998, Kenmore 2002). These combined beneﬁts should
encourage farmers to adopt IPM broadly, but the empirical evidence seems to indicate
that either extension systems have failed to communicate adequately all the potential
beneﬁts of IPM adoption or that adopting IPM has not provided all expected beneﬁts.

Farmer ﬁeld schools (FFS) have recently received major support from
development agencies as an alternative extension method that would improve the
delivery of IPM (F eder et al 2004a). By using participatory techniques and the “learning
by doing” approach F FS seek to combine the scientiﬁc knowledge and the practice of
IPM with farmers’ experience and interests, making these farmers the main actors of the
technology development (Gallagher 1998, Kenmore 2002). In addition, it is expected that
FFS graduates will convince other farmers about the value of IPM and increase its
diffusion (Fleischer et al 1999). The farmer based orientation of FF S is believed to be
especially suitable for dealing with the complexity of IPM and its combined potential
beneﬁts.

Economic impact analysis of IPM adoption and F F S implementation in
developing countries has mainly concentrated on changes on the level of knowledge

about IPM (Hussain et a1 1994, Godtland et al 2004, Feder et al 2004b) or on measuring

39

farm level effects (Walker & Crissman 1996, Feder et al 2004a, Swinton 2005). While
farmers’ knowledge of IPM usually increases after participating in IPM training
programs, studies of IPM proﬁtability have shown mixed results and no clear advantages
of IPM over chemical control options (Morse & Buhler 1997, Peder et al 2004a). In spite
of the growing recognition that there exist hidden costs related to the environmental and
health effects derived from pesticide use (Rola & Pingali 1993, Crissman et a1 1998 and
Maumbe & Swinton 2003), and that IPM adoption could help to reduce them, these costs
have been omitted from virtually all prior impact studies of IPM adoption. Incorporating
these environmental and health effects in the analysis could help to improve
understanding of farmers’ decisions about pest control.

This paper expands the traditional impact assessment of IPM delivery by adding
the E&H effects derived from its adoption. The recent implementation of farmer ﬁeld
schools (F FS) among Nicaraguan bean growers serves as a case study. Using farm level
data from a recent cross sectional survey among Nicaraguan bean growers, this paper ﬁrst
measures econometrically the standard economic impacts of farmer participation on IPM
training on pesticide use, adoption of IPM non-chemical practices and on bean yields and
net revenues. This farm level analysis distinguishes between F FS participation and
traditional IPM training programs, and it controls for the potential endogeneity associated
with participation in extension programs (Strauss et a1 1991, Owens et a] 2003, Feder et
al 2004a) using instrumental variables. It also tests whether the difﬁision effect of FFS
exists. Next, the paper incorporates the E&H impacts and measures the effect that F F S
and other IPM training programs have had on human health and environmental services.

Health impacts are represented by farmers’ self-reported acute illness symptoms, and

40

environmental impacts are represented by self-reported on-farm populations of beneﬁcial
insects.

Results highlight the importance of correcting for endogeneity in program
participation. The paper shows how failure to correct for endogeneity sharply exaggerates
beneﬁts from F FS participation. Not only did FFS produce no signiﬁcant gain in
graduates’ bean yields and incomes, FFS in Nicaragua failed to improve household health
and levels of beneﬁcial insects in comparison with alternative IPM training programs.
The paper discusses how farmers’ production conditions, the complexity of IPM and the
length of FFS training in Nicaragua can explain these results and identify more promising
IPM extension approaches.

2. Farmer ﬁeld schools and the delivery of IPM

Integrated Pest Management (IPM) is a group of pest management practices aimed to
reduce E&H risks using information about pest biology. One cornerstone principle is the
idea of keeping a crop pest infestation below an economic threshold level: the pest level
at which control measures are necessary to prevent decline in net returns (Fernandez-
Comejo et a1 1998, Bajwa and Kogan 2002). Pest control methods may include pesticides
when necessary, but also non-chemical inputs and specialized practices such as agro-
ecological analysiss, botanical insecticides, insect sticky traps and others (F emandez-
Comejo et al 1998).

The potential beneﬁts of IPM are expected to encourage farmers to adopt IPM
broadly. However, IPM is a complex knowledge-intensive management approach that

poses big challenges to the extension channels in charge of its diffusion (Sorby et a1

 

5 This practice consists in a permanent observation of the crop ﬁeld and the use of insect scouting as the
main activity. It will determine whether pest infestation has reached the economic threshold level.

41

2003). The delivery of IPM requires knowledgeable extensionists with sophisticated
communication skills (Lagnaoui et a1 2004).

Like most of extension efforts in developing countries, the delivery of IPM has
mainly followed the training and visit (T&V) approach (Benor et a1 1984), based on short
ﬁeld visits to selected farmers who are put in charge of delivering technical packages to
neighbor farmers (Piccioto & Anderson 1997). This extension approach has been
criticized for being “top down” and for failing to organize farmers adequately (Hussein et
al 1994). Training and visit methods are alleged not to have provided the farmer-based
approach required for dealing with the complexity of IPM and have failed to disseminate
all potential beneﬁts associated with IPM adoption (Cowan & Gunby 1996). The result
has been low adoption rates of IPM within the developing world (Addo et a1 2001,
Chaves & Riley 2001, Or 2003).

In the search for more suitable IPM extension approaches, Farmer Field Schools
(F F S) have received major support from development agencies that include the FAO and
the World Bank (Feder et al 2004a). FF S were ﬁrst established in Asia at the end of the
1980’s. FFS use participatory techniques and combine the scientiﬁc knowledge and the
practice of IPM with farmers’ experience and interests under the learning by doing
approach (Gallagher 1998). Within this new approach, farmers are believed to
incorporate their priorities, develop more skills and become the main actors in the
decision making process of the technology adoption (Kenmore 2002). In addition it is
expected that FFS graduates will be more able to convince other farmers about the IPM

value and increase its adoption (Fleischer et al 1999, Quizon et a1 2001).

42

Following reports of beneﬁcial impacts from the ﬁrst FFS attempts, F F S have
been implemented worldwide as the preferred method for delivering IPM. Unfortunately,
most of the positive, early assessments were not based in economic impact studies that
properly controlled for pre-existing differences among FFS graduates, other IPM training
program participants and farmers insulated from any IPM exposure (Feder et a1 2003,
Godtland et al 2004). The failure to address this potential endogeneity problem could
have caused extension impact estimates to be overestimated (Strauss et al 1991, Owens et
a1 2003). In addition, most economic impact studies of IPM delivery (F FS and otherwise)
have ignored the expected E & H impacts, concentrating instead on farmers’ knowledge
of IPM and farm level proﬁtability (Rola et a1 2002, Feder et al 2004a & 2004b, Godtland
et al 2004). This paper addresses both the endogeneity problem and the E&H impacts in a
comprehensive assessment of FFS among Nicaraguan bean farmers.

3. FFS experience among Nicaraguan bean growers

Nicaraguan agriculture is characterized by high agrochemical use compared to
most developing countries. It has one of the highest rates of pesticide consumption per
capita in the developing world with 1.5 Kg per person per year (PAHO 2002). The
relatively high pesticide consumption among small farmers has started to harm human
health and insect biodiversity in the area. A recent study estimates that 400,000 pesticide
poisonings may occur each year with 5% of people exposed to pesticides experiencing
illness symptoms (PANNA 2002). It is also reported that farmers in Central America
recognize that the overuse of pesticides is destroying the beneﬁcial insect population
(Bentley and Andrews 1996). These problems have led to a search for less harming

farming systems.

43

IPM training in Nicaragua is not new, and efforts to deliver this technology have
existed for the past 20 years. Following IPM initiatives that started in developed
countries, many local and regional research and development institutions have organized
IPM training programs in order to reduce the use of pesticides and to increase the
adoption of IPM practices (Bentley & Andrews 1996, Cobbe 1998, Staver & Guharay
2003). Unfortunately, up to 2001 adoption of IPM had not reached desired levels
(Bentley et a1 2001, PROMIPAC 2001).

In order to improve the adoption of IPM in Nicaragua, the Project for Integrated
Pest Management in Central America (PROMIPAC) has promoted the implementation of
Farmer Field Schools (F F S) since 2001.1n this country, F FS have been implemented by a
group of non-govemmental organizations (N GOs). Most of these NGOS had previous
experience in delivering IPM and have simultaneously kept using the training & visit
approach with other clients while implementing F PS. This feature allows a natural
comparison of FF S and T&V. All the NGOs had a common starting point in the FFS
implementation: sending extension agents to participate in an intensive training for
trainers program. In this program, extension agents learned their new role as facilitators
on the extension process, several participatory research techniques, and many new
scientiﬁc concepts related to IPM (PROMIPAC 2001). At the end of the training
program, facilitators were able to offer different alternatives to solve farmers problems,
especially ones related to pest control. Based on the training of trainers, differences in
curricula among individual FFS should only be attributed to different farmers’

preferences, not NGO implementation.

44

Nicaraguan FFS were developed to last for only one year conducting 13 weekly
sessions‘5 and developing a curricula that included IPM concepts (pest biology, diseases,
predators, biological control, etc), ﬁeld practices (mainly three IPM practices: Agro-
ecological analysis, botanical insecticides and yellow sticky traps) and on-farm group
experimentation (comparison of an IPM plot with a conventional crop plot, different level
of fertilization, alternative varieties resistant to diseases). There were no follow up
activities planned after FF S graduation. Before ﬁnalizing the training program, each FF S
organized open ﬁeld visits with community neighbors in order to share graduates’ new
knowledge and experience (PROMIPAC 2002, 2003). The rate of graduation reached
more than 80%, and more than 1000 farmers participated in Nicaraguan FFS between
2001 and 2003. Out of this total, around 200 farmers participated in the 13 bean FF S
(PROMIPAC 2001). According to preliminary evaluations, FFS graduates increased their
IPM knowledge by 40% (PROMIPAC 2001). However the impact of this knowledge
improvement on health and environmental outcomes remained unknown. The rest of the
paper conducts an expanded economic impact study.

4. The analytical framework

In order test the impacts of FPS and other IPM training programs on households’
outcomes, this paper presents a brief model of farm household behavior. Under the
agricultural production household approach (Singh et a1 1986) the following utility

function is assumed:

U=U(X,,X,, .X,.H) <1)

 

6 . . . . .
This is the average time for a crop season in Nicaragua

45

Each household derives utility from the consumption of farm goods (Xf), non-
farm goods (an), leisure (XI) and household health (H). Farm goods are produced

according to the following household production function:

QX. =QXa[D(P.Zc(K),ch(K))J(H),AE / CH] (2)

where

K=K(T/ CH)

The level of farm production (Qx,) depends on a damage function (D), the use of
labor (L), household assets (A) and environmental services (E). It is also inﬂuenced by
household characteristics and other shifters (CH). The damage ﬁrnction represents the
potential damage that pest and diseases can cause to farm production and depends on the
pest pressure (P), the use of chemical inputs (Zc) and the use of non-chemical inputs (ch)
like IPM practices. In addition, both the level of chemical (pesticide) and non—chemical
input use are determined by the knowledge (K) that farmers have about the pest and
diseases and about the available pest controls. This level of knowledge is inﬂuenced by
IPM training (T) and household characteristics (CH). It is expected that labor is increasing

in household health that has the following household production function:

H=H(H0,Xa,Zh,Zc/ CH) (3)

Household health has a natural status (Ho) that are determined by early nutrition,

health care and others factors not analyzed in this research. Health can be augmented by

46

consuming goods (especially food) and health inputs (Zh) like health care services and
protective devices, but health is assumed to be diminished by the use of chemical inputs
(H’(Zc) S 0). Environmental services (E) do not directly affect household utility, but they
beneﬁt household welfare by enhancing crop production. E is represented by beneﬁcial

insects and has the following production function

E=E(E0,Zc/ CH) (4)

Beneﬁcial insects have a natural endowment (E0) and are affected negatively by
chemical use (E’(Zc)5 0). The level of beneﬁcial insect population can be also affected
by household characteristics.

IPM training has an indirect effect on farm production through knowledge and
hence chemical and non-chemical input use. We can expect that Xa’(Zc) and Xa’(Zm) will
be positive because a greater use of both inputs will reduce crop damage and less crop
damage will increase farm production (positive cross partial derivatives). However, the
effect that better knowledge will have on chemical and non-chemical input use (Z¢’(K)
and ch’(K)) is unknown. Households with better knowledge about pests and controls will
know better when the pest pressure has reached the economic threshold and what is the
best control for that situation. But better information could imply higher use of chemical
and non-chemical inputs if the pest pressure becomes high. It could also mean less of
both inputs if the agro-ecosystem analysis results on a low requirement for control.
Similarly, the effect of IPM training on household health and beneﬁcial insect

populations is also analytically indeterminate.

47

Each IPM training method (T) has a different inﬂuence on the level of knowledge
about pests and their management. Greater exposure to IPM training will increase
participants’ knowledge (K’(T)>0). It is also expected that farmers exposed to IPM
training will disseminate that training via interaction with neighbors.

The household utility maximization problem is completed by incorporating the

household full income constraint:

Y=I:X.+w1. —e.Z.—Pz..2..—PZ.Z.-wz. <5)

where Pa, w, P10, Pch and P2}. are the exogenous prices for agricultural goods, labor,
chemical inputs, non-chemical inputs and health inputs. Solving the household utility
maximization problem we can derive the reduced form for the factor demand functions
for chemical inputs (2,) (pesticides) and non-chemical inputs (ch) (IPM practices), as
well as the household demand for health (H) and environmental services (E), and the
bean yield (On) and net revenue function (Yb)s based on some set of explanatory

variables:

B=K(B.,ch,PmB.M/1T,CH) (6)

where B is a vector representing Z6, ch, H, E, Ox, and Yb_
The main task in this paper is to measure econometrically the coefﬁcients of the T
vector that contains the estimated impacts of F F S, other IPM training methods and the

expected diffusion effect on the above input demand, output supply and net income

48

functions. Five testable hypotheses are related to the assumptions underlying most of
IPM delivery programs:
0 IPM training and especially FFS should:

a) Increase the adoption of IPM practices

b) Decrease pesticide use

c) Increase crop yields and net revenues.

0 Similarly it is also expected that

d) Farmers in direct contact with FFS graduates will experience similar effects.

e) FFS and other IPM training methods will improve H & E outcomes

(presumably a side effect of decreased pesticide use)
5. Empirical Analysis
5.1. Data
Farm level data were collected for this study between May and August 2004 via a cross-
sectional survey of 436 households of Nicaraguan bean growers in the departments of
Esteli, Matagalpa, Jinotega, Madriz, Masaya and Carazo. The sampling design followed a
double stratiﬁcation (Deaton 1997) based on exposure to IPM training in FF S and other
programs and covering diverse production settings in the main bean production areas.
Households were interviewed in 74 rural communities, including 13 where FFS were
implemented, 9 where F FS graduates lived but no F FS were held, 26 communities
selected randomly where no F FS existed but other IPM extension services were
available7, and 26 communities selected randomly where no IPM extension was present.

In each community, households were selected randomly and included both clients and

 

7 Other IPM extension services refers to the training and visit (T&V) approach that was mainly provided
by the same NGOs in charge of FFS implementation but usually targeted for NGOs clients who did not
participate in FFS.

49

non-clients of NGOs. The ﬁnal sample included FF S graduates, farmers participating in
other IPM programs, farmers who experienced double IPM training (by both FFS and
other means), neighbor farmers of FFS graduates and farmers with no prior contact with
formal extension.

5.2. Potential econometric problems

The main potential econometric problem of this research is endogeneity
associated with self selection of farmer participants in IPM extension programs and the
non-random placement of these programs. Many studies have recognized that failure to
correct this problem produces biased estimates of the extension service impact, that
usually result in overestimates (Strauss et a1 1991, Owens et a1 2003, F eder et al 2004a).

The F F8 in Nicaragua were located at sites with good access for meetings and
experimentation that could capitalize on pre-existing farmer-NGO relationships. Farmers
with good pre-existing relationships with NGOs could have been more willing to
participate in IPM training whereas those with poor NGO relationships could have been
less so.

This paper evaluates ﬁrst the convenience of using two stage least squares (2SLS)
estimation to correct selectivity bias in the econometric estimation, which is the preferred
method in linear models with cross sectional data (Wooldridge 2002a). Some models in
this paper have a clear non-linear nature and would be more suitable for a trivariate
probit, ordered probit or multinomial logit speciﬁcation, but correcting selectivity bias in
non-linear models through the recommended “control function approach” generates
computational problems that can only be applied under normality conditions (Wooldridge

2002a, Wooldridge 2002b, Blundell & Powell 2004). But as Angrist (2001) pointed out,

50

standard instrumental variable estimation can still be applied to non-linear models when
the objective is to estimate the causal effect of the endogenous variables. Wooldridge
(1997) also showed that ZSLS could still be applied in a random coefﬁcient model with a
binary endogenous variable when this variable follows a linear probability model, the
linear approximation of models with binary dependent variables as the adoption of IPM
practices case. Even recognizing the non-linear nature of some models, this paper will
mainly use ZSLS procedures. Only when exogeneity cannot be rejected in some linear
approximations, this paper prefers non-linear models.

The sample design poses secondary econometric problems of a clustered and
stratiﬁed sample that can bias the parameter estimates (Wooldridge 2002a). To deal with
the clustering problem, this paper adjusts the variance matrix and includes dummy
variables for each cluster using survey regression methods (Wooldridge 2003). This
method is suitable given the nature of the sample design that has small clusters (six
households per village on average) and a large number of clusters (74 villages). To
correct the unbalanced representation of farmers (especially FFS trainees) in the sample,
this paper uses a weighting scheme as suggested by Wooldridge (2002a) to adjust results
to the population of bean all bean growers (see Appendix A3 for details).

5.3. The econometric speciﬁcation

This paper speciﬁes three linear models for pesticide demand, one linear model
and three linear probability models for the adoption of IPM practices, one linear model
for the bean yield function, one linear model for the bean net revenue ﬁrnction, four
ordered probit models for the household health demand and one ordered probit and one

probit for the household demand for environmental services. All linear speciﬁcations are

51

estimated by 2SLS. The non-linear models are used when the exogeneity could not be
rejected in the linear approximations of these models.
All models have the same set of explanatory variables, following the general

model structure:

DJ =PK,BP+TT,BT+CH,BCH+CC,BCC+U ;F0rj=1,--.13 (7)

Each dependent variable (Zc, ch, H, E, Qxa and Yb) depends on vectors of K
output and input prices (PK), household participation in the T IPM training program (Tr),
socioeconomic and other household characteristics (CH) and community ﬁxed effects
(Cc), with disturbances assumed to be independently distributed (Ui). All variables used
in the econometric estimation are described in detail in the following subsections.

This paper uses the predicted probability of each IPM training program
participation (F F S and other programs) as generated instruments of the true program
participation in all the ZSLS models. Each predicted probability is estimated using a
probit speciﬁcation that has as explanatory variables the original exogenous variables and
some redundant variables that explain the variation of FFS and other IPM program
participation (Wooldridge 2002a). The structure of each probit speciﬁcation uses the
same set of explanatory variables as the previous models plus a vector of Z variables
containing the redundant variables related to participation in IPM training, but believed to
be uncorrelated with the disturbance term of the IPM adoption, pesticide demand, bean

yields, net revenue and E&H outcomes equations (see second table of Appendix A8).

52

5.3.1. The dependent variables

Household input demand for insecticides, herbicides and fungicides is represented
by the quantity of pesticide active ingredients used by each household in bean production
during the most recent season in 2003 weighted by acute human toxicity. A human
toxicity index is calculated for each household i as the sum over all m pesticide active
ingredients (aiik in kg/ha, k=1,. . .,k) divided by each active ingredient’s mammalian
toxicity (measured by the minimum dose per gram of body weight that is lethal to 50%
of a test rat population or LD50, as reported in USDA, 1998). The human toxicity index,

show below, is proportional to the LDSO and is increasing in lethality.

 

ai.
HTI. = E "‘
' L050”,

The adoption of the three speciﬁc IPM practices promoted in FFS is represented
by binary variables that indicate whether a household adopted agro-ecological analysis,
botanical insecticides and/or yellow insect traps and a variable that measures whether a
household adopted one, two or the three IPM practices. As explained before this paper
uses a linear approximation for these non-linear models.

Health outcomes are measured as changes in the incidence of four acute illness
symptoms among household members (respiratory difﬁculties, eye irritation, stomach
ache and blurred visions). Each dependent variable measures whether a household

experienced a decrease, no change or an increase (-1, 0, 1) in the incidence of each acute

 

’ Other models were considered for changes in the number of acute symptoms and for changes in the
incidence of diarrhea, headache, skin rash, dizziness and muscle pain but the null hypothesis of the joint
signiﬁcance of the overall model could not be rejected.

53

symptom during the most recent season in 2003 compared to average symptom incidence
prior to FF S participation or contact.

The beneﬁcial insect outcomes are represented by two responses. The ﬁrst was a
3 point Likert scale of on-farm beneﬁcial insect population levels during the most recent
bean season9. The second was whether farmers reported having observed an adequate
level of beneﬁcial insects for controlling, at least partially, their pest problems in beans.

As the beneﬁcial insect observation was restricted to farmers participating in F F S
and other IPM training programs, the sample size is reduced in both models to 144 and
212 observations respectively.

Bean yield is a continuous variable measured in kg/ha during the most recent
season in 2003. Bean net revenue is an expected measure in US$/ha based on farmers’
self-reported revenues and cost for the same bean season.

5.3.2. The explanatory variables

Output prices are selling prices for bean and maize reported in USS/kg. Input
prices include bean seed, metamidophos insecticide, gramoxone (paraquat) herbicide and
mancozeb fungicide"). The bean yield and net revenue functions also included the price
of fertilizers”. Other prices and price proxies are the transport cost to the nearest market
of a sack of beans or maize and the distance between each farm and the municipal capital.
Other production variables include the most recent farming season that bean were grown:

primera, postrera or apante, the use of disease resistant bean varieties, altitude (proxy for

 

9 The main beneﬁcial insects related to bean production in Nicaragua are: Eretmocerus serious, Chysopa
oculata, l-Iippodamia convergens and Encarsia spp.

'0 Metamidophos, paraquat and mancozeb were by far the main active ingredients among insecticides,
herbicides and fungicides on the Nicaraguan farms surveyed in 2004

" we use the price of the most common combination of nitrogen, phosphorus and potassium (12-30-10)

54

rainfall and temperature) and farmers’ observed level of bean pests and diseases (high or
not).

IPM training participation is speciﬁed for F FS graduates, other IPM training
program participants and the interaction of the two”. Neighboring farmers inﬂuenced by
FFS graduates are also speciﬁed, and households without IPM training contact are kept as
the control group. It is also included the number of years after FFS graduation. Given that
each FFS took place during only one year, this variable measures the years that farmers
had for implementing what they learned after FF S graduation.

Household characteristics are represented by two groups of variables: (1)
measures of household assets: farm size, area under irrigation, number of cattle, whether
family has electricity at home, and whether the household receives remittances from
relatives working in foreign countries; and (2) measures of household composition:
gender of household head, age and years of education of household head, percentage of
females, and percentage of family members under 14 years old. Other conditioning
variables include whether the household relied upon hired labor or family labor for
pesticide application and whether household members used protective devices like mask,
gloves or special clothing while applying chemicals. Dummy variables are included for
community-level ﬁxed effects, with one of them kept as a control. The 73 community
dummies are used for the ZSLS estimations, but are substituted by municipality dummy
variables (9 out of the 10 municipalities) in the estimation of the predicted probability of
F FS and other IPM training participation.

Finally, redundant variables are used to control for endogeneity of extension

participation. All these variables measure pre-existing linkages between individual NGO

 

'2 There were 35 FFS graduates who had previous participation in other IPM training programs

55

and client farmers, including whether households received previous credits from the
NGO, whether they received food assistance or any cash support, and whether they
adopted soil conservation practices prior to their IPM training (terraces, contour lines and
live fences). As shown in Appendix A.8, many of these redundant variables meet the ﬁrst
condition for a good instrument by signiﬁcantly explaining farmer participation in FF S
and other training programs. These variables seem to also meet the second condition for a '
good instrument: they not be correlated with the variables that explain part of the
variability of the household demand models but were not observed. Credits and other
NGO support received by farmers prior to the most recent bean season should not
inﬂuence their decisions about the level of pesticide use or the adoption of IPM practices
during the season. Neither should E&H effects derived from pesticide use decisions be
correlated with redundant variables that are unrelated to the most recent bean season. Nor
should prior adoption of soil conservation practices condition farmers’ decisions about
pesticides and non-chemical practices used during the most recent bean season. The
adoption of terraces, contour lines and live fences are not related to the levels of pests and
diseases or the associated pest control measures. The characteristics of these variables
make them suitable for specifying robust IV models that will be tested for correcting
endogeneity of farmer participation in FPS and other IPM programs.

6. Results

Table 2.1 presents the main differences in households’ farm, health and environmental
outcomes among the ﬁve different groups of farmers according to their exposure to [PM
training. Although less pesticide use is expected among farmers exposed to IPM training,

only FFS graduates reported a lower proportion of insecticide users (even lower for FFS

56

Table 2.1. Differences in pesticide use, adoption of IPM practices and bean yields
and net revenues among ﬁve groups of Nicaraguan bean growers during 2003 bean

 

 

season.
FFS Other 1PM FFS&IPM Inﬂuenced Insulated

Number of farmers 90 129 35 51 131
Pesticide user on beans

Insecticides (%) 66 82" 54" 80* 78*

Herbicides (%) 56 50 43" 61 64

Fungicides (%) 19 29" 34*" 17 12"
IPM adoption (%)

Agro-ecological Analysis 61 32* " 69 2* * * 1" ‘

Botanical Insecticides 56 55 83‘”I 3"" 2""

Insect traps 30 22 48" 1*" 1‘"
Bean yields (kg) 708 678 916” 884‘ 637‘“
Bean net revenues (U SS) 132 120 190" 190" 109

Gross revenues 240 233 304" 274 209'

Total cost 108 113 1 14 84* 100

Pesticide cost 16 18 10" 17 19

Farmers with losses (%) 24 26 17" 10"" 17"
Changes health outcomes(%)

Decreased resp. difﬁculties 5.6 6.7“ 5.7 9.7“ 8.4”

Increased resp. difﬁculties 0.0 0.7 0.0 0.0 0.8

Decreased eye irritation 15.6 289‘" 429‘" 22.6" 28.2’"

Increased eye irritation 1.1 0.7 0.0 0.0 0.0

Decreased stomachache 13.3 16.1" 1 1.4 19.4" 13.7

Increased stomachache 0.0 0 0.0 0.0 0.8

Decreased blurred vision 10.0 22.1" 17.1“ 355“" 18.3"l

Increased blurred vision 2.2 2.7 2.9 0.0 0.8
Changes beneﬁcial insects+

Observed higher level 35.1 52.6 67.9

Observed lower level 45.6 15.8 14.3

Adequate level 21 .1 4.7 37. 1

 

Signiﬁcantly different from FFS at 10% (‘), 5% C") and 1% (*") levels using x2 test.
+ Only farmers exposed to IPM training observed beneﬁcial insects.

 

graduates who had participated previously in other IPM training programs). However,

farmers exposed to T&V IPM extension and F FS neighbor had the greatest proportion of

57

insecticide users. There were very few signiﬁcant differences across groups of farmers in
the proportion of herbicide users; only a double (F FS plus other) IPM training reduced
the proportion of users. Finally, fungicide users were unexpectedly more common among
participants in any IPM training program. This could have resulted from IPM training
improving farmers’ ability to determine the economic threshold for disease control, so it
is not necessarily a bad outcome.

The adoption of IPM practices is restricted almost exclusively to farmers directly
exposed to IPM training. There was no spillover effect of FF S to farmer neighbors of FFS
graduates. FFS showed no signiﬁcant advantage over the T&V approach at inducing
greater adoption of IPM practices. F FS were more likely to adopt agro-ecological
analysis, but no more likely to adopt botanical insecticides and insect sticky traps than
other IPM-exposed farmers (Table 2.1).

Bean yields and net revenues were signiﬁcantly higher among farmers exposed to
double IPM training and among neighbor farmers expected to be inﬂuenced by FFS
graduates. These results are another reason for testing a potential endogeneity problem in
FFS participation. More productive and proﬁtable farmers could have been excluded
from FFS or could have not elected to participate.

Health outcomes improved signiﬁcantly among all groups, but F FS graduates
experienced the lowest improvement in the incidence of the four acute health symptoms
associated with pesticide use (Table 2.1). Even farmers with no IPM training
participation had better health outcomes during the 2003 bean season closest to the early
2004 farm survey. Farmers who participated in other IPM training programs or

experienced both FF S and other IPM training had a greater reduction in the incidence of

58

eye irritation (Table 2.1). Environmental outcomes were observable only among farmers
who had attended IPM training programs. Forty six percent of F FS graduates observed a
lower level of beneﬁcial insects on their farms in comparison with the most recent 2003
bean season while 35% observed higher level of these beneﬁcial insects. Other IPM
training participants and farmers who received a double IPM training observed more
beneﬁcial insects.

Table 2.2. Descriptive statistics for selected variables used in the survey least
squares for 436 Nicaraﬂan bean growers, 2003-04+

 

 

 

Variable description Mean Std Dev.

Dependent Variables

Index of insecticides weighted by toxicity (x 100) 2.9 5.1

Index of herbicides weighted by toxicity (x 100) 0.5 0.7

Index of fungicides weighted by toxicity (x 100) 0.1 0.6

Adopted agro-ecological analysis (%) 28.2

Adopted botanical insecticides (%) 35.1

Adopted insect yellow traps (%) .1 5.8

Adopted only one IPM practice (%) 19.1

Adopted two IPM practices (%) 14.9

Adopted three IPM practices (%) 10.8

Bean yields (kg/ha) 688.0 578.0

Bean net revenue (U S$/ha) 91.2 124.9

Changes in Acute illness symptoms
Number -1.8 2.4
Respiratory difﬁculties (-1,0, 1) -0.1 0.3
Eye irritation (-1,0,1) -0.3 0.5
Stomachache (-1,0,1) -0.1 0.4
Blurred vision (-1,0, 1) —0.2 0.4

Level of beneﬁcial insects (1,2,3) (N=144) 2.1 0.9

Observed enough beneﬁcials (%) 9.0

[PM training

FFS participation (%) 20-6

Other IPM training participation (%) 34-2

FFS and other IPM participation (%) 8.0

Inﬂuenced by FFS graduates 12-0

Years after FFS graduation 2-3 1-7

 

+ Complete list of variables in Table A3] (appendix)
Descriptive statistics for variables in the econometric analysis are summarized in
Table 2.2, while selected regression results related to the impact of FF S and other IPM

training programs on farm, health and environmental outcomes are showed in tables 2.3,

59

2.4, 2.5, 2.6 and 2.7. Hausman tests were performed in all models in order to test the need
for endogeneity correction in 2SLS models. The three pesticide demand models and both
the bean yield and net revenue models have a clearly linear nature, so using 2SLS seems
to be the most appropriate to control for endogeneity of training program participation.
In most of these models, the exogeneity hypothesis was rejected (Tables 2.3 & 2.5)
suggesting that the IV estimates are more suitable. Only in the fungicide demand model
could the exogeneity hypothesis not be rejected (Table 2.3), making the OLS estimates
preferred in this case.

The adoption of IPM practices, the changes in the incidence of acute illness
symptoms and the level of beneﬁcial insects are integer dependent variable models that
are clearly non-linear. However, after performing Hausman tests to their linear
approximation, the null hypothesis of exogeneity could not be rejected in any of the
health and environmental models (See appendix A8 for details), so ordered probit and
probit speciﬁcations are used. Two 2SLS models are used to estimate IPM adoption due
to evidence of endogeneity in all adoption decisions except insect yellow traps (Table
2.4).

Econometric results show that the assumptions underlying most IPM programs
did not hold for FFS. F FS participation had no signiﬁcant impacts on pesticide use
(hypothesis a), adoption of IPM practices (hypothesis b), bean yield or bean net revenue
(hypothesis 0) (Tables 2.3, 2.4 and 2.5). There was strong evidence of endogeneity in
FFS participation and failing to control for this problem provides erroneous estimates of
FFS impacts. Like many previous studies, the OLS estimates indicate that FF S graduates

adopted more IPM practices than farmers without IPM training, increased the probability

60

 

mix—£8 _8_wo_oooio&a mo axiom 5a.: 05 mm was—88 885 I

223530 >_ 05 .5.“ ESE $83ch 05 Ext—:25 .w.< Season—am E 5597. 8a 3.38.. c2329. 82956 +
885533 E norms—3m a mo 2:? 829.3"

a: a “some”; t... can a Esme»; .. ”as a “acacia ..

 

 

 

an; ~23. 885 m :55 as? a
an... an... M: So oars... a
no. .8835:

83: 3.8 A38 2 _. c 93.8 80.8 ans 5.8
83.. 3.3. w: .? 83 83. 83. Ed. 83. pa» coco no;

$2: an : A38 Ewe 3.3 :3: 5.8 5. c
23.. $8.? ~85 $3- 83 Rod. 33. .23. 883:5

no.8 5.8 A88 :3: see 5.: : _.: 5.8
83 :3 23 89° 82 $3. 82 2 :7 2.: a. m":

:98 53 34.0 5.5 $3 and 5.0 853
353.46 at. _v_.c 3.1.36 neawamd ecgmd ﬁlmm—d :eowNA enemawd Zn: .550

$4.8 an. 5 A38 :56 $2: Gad awe 8&0
8:. £3 :2 .52 and 5.355 82 rag; ma...

8203035 323835 80:85 80:85
mag 89.:— >~ want 885 $8283 2 Eggom wccsoom Z Iwczsoom «0 595.2 >~ mo cons—=2

+3483 .9539. .32. an «.332

wmv .5. 3.53.. comma—we.— mouasum ~23— motsm ASE—ecu SE no ”ﬂaws... 2.: .55: van mhh we Babe 2: .QN 93...”.

 

61

of adopting insect scouting and botanical insecticides, and reduced slightly the toxicity-

weighted insecticide use.

These results on F FS impacts in Nicaragua are consistent with Feder et al’s

(2004a) ﬁndings among Indonesian FFS graduates, but not all their explanations about
FFS poor performance are applicable to the Nicaraguan case. One possible cause of the
failure of Indonesian FF S was complexity of the IPM knowledge gained (Feder et al.,
2004). However, in the Nicaraguan case, participants in the “other IPM” training
(following T & V approach) were able to signiﬁcantly increase the likelihood of adopting
insect scouting by 38%, botanical insecticides by 65% and insect scouting by 25%.

Table 2.4. The effect of FFS and other IPM training on pesticide use. Survey least
squares regression results for 436 Nicaraguan bean growers, 2003-04+

 

 

Insecticide IV Insect. Herbicides IV Herb. Fungicide 1V FunL
F FS -0.027* * -0. 106 0.003 -0.007 0.001 0.001
(2.38) (0.90) (1.42) (0.65) (0.75) (0.29)
Other IPM 0.004 -0.002 0.001 0.001 0.000 0.002
(0.53) (0.09) (1.47) (0.84) (0.58) (1.20)
FF S & IPM 0.007 -0.156 -0.007"‘ -0.026"' -0.001 -0.001
(0.54) (1.04) (1.94) (1.66) (0.80) (0.17)
Years aﬁer grad. 0.007 0.042 -0.002"' 0.003 -0.000 0.000
(1.12) (0.93) (1.86) (0.77) (0.32) (0.25)
Inﬂuenced 0.003 -0.017 0.001 -0.001 -0.001 -0.000
(0.53) (0.83) (0.94) (0.33) (0.90) (0.06)
Hausman test
F statistic 4.36 2.01 1.46
P value 0.0024 0.0971 0.2179

 

" Signiﬁcant at 10%; " signiﬁcant at 5%; "”” signiﬁcant at 1%

Absolute value of t statistics in parentheses
+ Complete regression results are shown in appendix A.8.

 

Hence, complexity of IPM methods cannot explain the lack of beneﬁcial impacts among

FFS graduates. Unfortunately, farmers trained ﬁrst in IPM through T&V did not improve

their farm outcomes after repeating the IPM training through F FS (Tables 2.3, 2.4 & 2.5)

62

On the other hand, poor production conditions among small farmers related to soil
and pest levels could reduce impacts of IPM extension in ways that could be difﬁcult to
measure through econometric analysis in Indonesia and Nicaragua (Feder et al 2004a). In
the Nicaragua case, this can be observed among farmers who participated in both FF S
and other IPM training. Although these farmers had the greatest bean yields and net
revenues among all the interviewed farmers (Table 2.1), the IV regression results indicate
that participating in both types of IPM training has no effect on bean yields and net
revenues. If this happens with the best group of farmers, it is very likely that some
variables related to production characteristics that are not observed by the analyst could
explain these outcomes.

Table 2.5. The effect of FFS and other IPM training on bean yields and net revenue.
Survey least squares regession results for 436 Nicaraggan bean growers, 2003-04+

 

 

 

Yield IV Yield Net revenue IV Net revenue
F FS -0.439 -19.670 264.924 -276.0538
(0.19) (1.25) (0.43) (0.99)
Other IPM -0.577 -2.650 -191.449 21.22
(0.56) (0.74) (0.70) (0.36)
FFS & 1PM 1.234 0.715 -23.459 33.91
(0.44) (0.005) (0.03) (0.09)
Inﬂuenced 2.068 0.980 627.566 42.602
(1.01) (0.32) (1.17) (0.87)
Years aﬁer graduation 0.114 8.983 -114.307 136.4294
(0.1 1) (1.44) (0.41) (1.25)
Hausman test
F statistic 3.29 2.59
P value 0.0132 0.0396

 

‘ Signiﬁcant at 10%; *"' signiﬁcant at 5%; "* signiﬁcant at 1%
Absolute value of t statistics in parentheses
+ Complete regression results are shown in appendix A.8.

 

Hypothesis ((1) seeks to test for the spillover effects of FFS implementation. Our

results strongly support F eder et al’s (2004a) ﬁndings FFS graduates did not produce any

63

signiﬁcant inﬂuence on their neighbors’ IPM adoption or performance (Tables 2.3, 2.4 &
2.5).

Hypothesis (e) tests for the E & H effects that are expected to bring extra beneﬁts
to FFS graduates and other farmers exposed to IPM training. Although all groups of
farmers experienced decreased incidence of acute illness symptoms during the most
recent 2003 bean season compared to previous years (Table 2.1), FFS graduates gained
less. In Table 2.6, the positive coefﬁcients for changes in the incidence of respiratory
difﬁculties and eye irritation do not imply an increase on the incidence of these
symptoms, but they do imply that F F S participation induced a lower improvement in
these health outcomes. This effect was partially offset by farmers with more years after
FFS graduation in the instance of respiratory difﬁculty (Table 2.6). In addition, farmers
who were ﬁrst trained in IPM through T&V and then graduated from a FFS reduced the
incidence of respiratory difficulties and eye irritation.

Table 2.6. The effect of FFS and other IPM training on household health outcomes.

Survey ordered probit regression results for 436 Nicaraguan bean growers, 2003-
04+

 

 

Respiratory difﬁc. Eye irritation Stomachache Blurred vision
FFS 1.137" 1.002“ 0.156 .0021
(2.18) (1.89) (0.28) (0.04)
Other IPM 0.009 0.218 -0.092 0.155
(0.03) (1.15) (0.42) (0.75)
FFS & IPM -0.950" -1.204* -0.082 0.120
(1.69) (1.92) (0.13) (0.19)
Inﬂuenced -0.341 0.636" -0.276 -0.332
(0.84) (2.30) (0.83) (1.17)
Years after graduation -0.452“ -0.156 -0.185 0.230
(1.69) (0.64) (0.70) (0.95)

 

‘ Signiﬁcant at 10%; " signiﬁcant at 5%; "* signiﬁcant at 1%
Absolute value of t statistics in parentheses
+ Complete regression results are shown in appendix A.8

 

64

F F S also performed poorly on environmental outcomes. FF S alone do not have
any signiﬁcant effect on the level of beneﬁcial insects by either measure (Table 2.7).
Farmers who had attended other IPM extension programs observed greater levels of
beneﬁcial insects and were more likely to report an adequate level of beneﬁcial insects to
control bean pests. This group of farmers did not improve their environmental outcomes
after receiving a second IPM training through FFS.
Table 2.7. The effect of FFS and other IPM training on environmental outcomes.

Survey ordered probit and probit regression results for 436 Nicaraguan bean
_growers, 2003-04+

 

 

Adequate level of beneﬁcials Level of beneﬁcial insects
N=212 N=l44
FF 8 0.747 -0.187
(0.98) (0.23)
Other IPM 0.873" 1864*"
(2.35) (4.91)
FFS & IPM 0.820 -0.281
(0.99) (0.36)
Years after graduation 0.464 0.492
(1.35) (1.35)

 

“ Signiﬁcant at 10%; " signiﬁcant at 5%; "" signiﬁcant at 1%
Absolute value of t statistics in parentheses
+ Complete regression results are shown in appendix A.8.

 

The inferior impact of the Nicaraguan F FS in comparison with other IPM training
approaches can also be explained in part by the length of exposure to IPM training. While
T&V extension used to keep their IPM programs for many years with most of the same
farmers, F F8 in Nicaragua lasted only one year with no follow up activities.

7. Conclusions

Impact analysis of educational activities requires accounting for pre-training
differences among participants and among the control or comparison group. The
Nicaraguan F F S experience shows that failing to correct for endogeneity associated with

IPM training participation created a misleading impression of beneﬁcial FFS impacts.

65

OLS regression estimates suggest that FF S graduates were more likely to adopt IPM
practices and reduced the toxicity of their insecticide and herbicide use . However, after
controlling for the endogeneity of farmer program participation through instrumental
variable estimation, the results show no signiﬁcant impacts of FFS among Nicaraguan
bean growers.

FFS have received growing support from development agencies for delivering
IPM, in spite of their high costs compared to other IPM training programs (Quizon et al
2001, Feder et al 2004b). Aid agencies usually expect that FFS will provide their
graduates with greater farm, health and environmental beneﬁts than the traditional
Training and Visit (T&V) extension approach. Our results do not support these
expectations. Instead, they show inferior performance of FFS compared to other IPM
training programs in Nicaragua. Moreover, there was no evidence of an FFS diffusion
effect among the neighbors of FFS graduates.

Why did F FS fail to make a difference? First, as F eder et al (2004a) found in
Indonesia, the poor soil and high pest infestation levels associated with small farmers in
developing countries could have limited the potential impacts of F F S. Even the group of
surveyed farmers with the highest bean yields and net revenues had lower yields and
incomes than average Nicaraguan levels in 2003 (FAOSTAT). Second, contrary to F eder
et al’s (2004a) claim that the complexity of IPM strongly impedes its delivery, the length
of the implementation of IPM training programs seems to be more a relevant explanatory
factor. In Nicaragua, bean farmers trained in IPM through the T&V approach increased
their adoption of non-chemical IPM practices whereas FFS training had no effect. The

main difference is that Nicaraguan F FS training lasted only one year, whereas the T& V

66

IPM training although less intensive was more continuous resulting in better farm, health
and environmental outcomes.

Given the high cost of delivering IPM through FFS (Quizon et al 2001, F eder et a1
2004bthe F FS extension approach should be reconsidered. This paper’s ﬁndings support
the recommendation of cost cutting and curricula prioritization in FFS programs (Feder et
al 2004a). But they also suggest a need to prolong IPM training programs in order to
guarantee more interaction between farmers and facilitators and an improved feedback
process. Longer IPM training should not mean repeating the current costly FFS over
many years. Rather the redesigned FFS should take advantage of cost savings from
reduction and prioritization of FF S curricula while maintaining the active participation of
farmers. Less investment per year could allow longer training without increasing
extension costs. The combination of these two changes could overcome potential
difﬁculties posed by the complexity of IPM and its delivery.

Although FFS did not provide better environmental and health beneﬁts to their
graduates, it is important to include these beneﬁts in the economic impact analysis of
IPM extension. This paper demonstrates that under the right conditions, the delivery of
IPM can increase a farm’s level of beneﬁcial insects, which could help to control crop
pests and eventually reduce the use of pesticides. Given the unknown direction of
pesticide use under many IPM approaches, it is important to measure health outcomes.
Farmer health should be improved by IPM training but as the Nicaraguan case shows,
exposure to IPM training can sometimes result in negative health outcomes.

Future research is needed in order to improve the methods of incorporating E&H

effects into an economic impact analysis. In addition to more accurate measures of acute

67

health symptoms and beneﬁcial insect populations (with the help of scientists from other
disciplines), it would be desirable to include long-term health outcomes and off-farm
environmental effects that are usually not observed by farmers who make pesticide use

decisions.

68

References

Addo, S., L. Birkinshaw and R. Hodges (2002). Ten years after the arrival in Ghana of
Larger Grain Borer: farmers' responses and adoption of IPM strategies.
International Journal of Pest Management 48(4): 315-0325.

Angrist, J. (2001). Estimations of limited dependent variable models with dummy
endogenous regressors: Simple strategies for empirical practice. Journal of
Business and Economic Statistics 19:2-16

Bajwa, W. and M. Kogan (2002). Compendium of IPM deﬁnitions (CID). What is IPM
and how is it Deﬁned in Worldwide Literature? IPPC publication N. 998.
Integrated Plant protection Center (IPPC). Oregon State University. Corvallis, OR
97331, USA.

Benor, D.; J. Harrison and M. Baxter (1984). Agricultural Extension: the Training and
Visit System. World Bank, Washington.

Bentley, J. and K. Andrews (1996). Through the roadblocks: IPM and Central American
smallholders. International Institute for Environment and Development.
Gatekeeper series N. 56.

Bentley, J ., L. Morales and A. Zamora (2001). La Extension de MIP en Nicaragua.
CATIE, Managua. 42pp.

Blundell, R and J. Powell (2004). Endogeneity in semiparametric binary response
models. Review of Economic Studies 71: 655-679.

Chaves, B. and J. Riley (2001). Determination of factors inﬂuencing integrated pest
management in coffee berry borer in Colombian farms. Agriculture, Ecosystems
and Environment 87:1 59-177.

Cobbe, R. (1998). Capacitacién Participativa en el Manejo Integrado de Plagas (MIP):
Una Propuesta para América Latina. F AO, Rome, 46pp.

Cowan, R. and P. Gunby (1996) Sprayed to death: path dependence, lock—in and pest
control strategies. Economic Journal 106:521-542.

Crissman, C., J. Antle and S. Capalbo (1998). Economic, Environmental and Health
Tradeoﬂs in Agriculture: Pesticides and the Sustainability of Andean Potato
Production. Dordrecht: Kluwer Academic Publishers.

Deaton, A. (1997). The Analysis of Household Surveys. A Micro Econometric Approach

to Development Policy. The John Hopkins University Press, Baltimore and
London.

69

Feder, G., R. Murgai and J. Quizon (2004a). Sending farmers back to school: The impact
of farmer ﬁeld schools in Indonesia. Review of A gricultural Economics. 26(1):
45-62.

Feder, G., R. Murgai and J. Quizon (2004b). The acquisition and diffusion of knowledge:
the case of pest management training in farmer ﬁeld schools, Indonesia. Journal .
of Agricultural Economics 55(2): 217-239.

Femandez-Comejo, J ., S. Jans and M. Smith (1998). Issues in the economics of pesticide
use in agriculture: a review of the empirical evidence. Review of Agricultural
Economics 20(2): 462-488.

Fleisher, G., F. Jungluth, H. Waibel and J .C. Zadoks (1999). A ﬁeld practitioner’s guide
to economic evaluation of IPM. Pesticide Policy Project Publication Series No 9.
University of Hannover, Germany.

Gallagher, K (1998). Farmer Field Schools (F FS): A Group Extension Process Based on
an Adult Non-Formal Education Methods. Rome, F AO, Global IPM facility.

Godtland, E., E. Sadoulet, A. DeJanvry, R. Murgai and O. Ortiz (2003). Impact of farmer
ﬁeld schools on knowledge and productivity: A study of potato farmers in the
Peruvian Andes. Economic Development and Cultural Change 53(1):63-92.

Hussain, S., D. Byerlee and P. Heisey (1994). Impacts of the training and visit extension
system on farmers’ knowledge and adoption of technology: evidence from
Pakistan. Agricultural Economics 10: 39-47.

Kenmore, P. (2002). Integrated Pest Management. International Journal of Occupational
& Environmental Health 8(3):]73-174.

Maumbe, B. and S. Swinton (2003). Hidden health costs of pesticide use in Zimbabwe’s
smallholder cotton growers. Social Science and Medicine 57: 1559-1571.

Lagnaoui, A., E. Santi and F. Santucci (2004). Strategic communication for integrated
pest management. Paper presented at annual conference of the International
Association for Impact Assessment, Vancouver, Canada.

Morse, S. and W. Buhler (1997). Integrated Pest Management. Ideals and Realities in
Developing Countries. Boulder, CO: Lynne Rienner, 200 pp.

Orr, A. (2003). IPM for resource poor African farmers. Is the emperor naked? World
Development 31(5): 831-845.

Owens, T., J. Hoddinott and B. Kinsey (2003). The impact of agricultural extension on

farm production in resettlement areas of Zimbabwe. Economic Development and
Cultural Change 51(2): 337-357.

70

PAHO (2002). Pan-American Health Organization. Epidemiological Bulletin 23 Num 3.

PANNA (2002). Pesticide Action Network North America. Pesticide Poisoning-a Central
American Epidemic. Global Pesticide Campaigner Vol 12 num 3.

Piccioto, R. and J. Anderson (1997). Reconsidering agricultural extension. World Bank
Research Observer. 12(2):249—259.

PROMIPAC (Proyecto Manejo Integrado de Plagas en America Central) (2001) -
Evaluacion de Metas del Proyecto PROMIPAC en su Segunda Fase 1998-2001.
Esteli, Nicaragua.

PROMIPAC (Proyecto Manejo Integrado de Plagas en America Central) (2002-3)
Informes de las Escuelas de Campo de Agricultores de Frijol. Esteli, Nicaragua.

Quizon, J ., G. F eder and R. Murgai (2001) Fiscal sustainability of agricultural extension:
the case of the farmer ﬁeld school approach. Journal of International Agricultural
Extension and Education 8: 13-24.

Rola A and P. Pingali (1993). Pesticides, Rice Productivity, and Farmer ’s Health: An
Economic Assessment. International Rice Research Institute and World Resources
Institute, Manila.

Rola, A., S. Jarnias and J. Quizon (2002). Do Farmer Field School Graduates Retain and
Share What They Learn? An Investigation in Iloilo, Philippines. Journal of
International Agricultural and Extension Education 9(1): 65-76.

Singh, 1., L. Squire and J. Strauss, eds (1986). Agricultural Household Models:
Extensions, Applications and Policy. Johns Hopkins University Press, Baltimore.

Sorby, K., G. Fleisher and E. Pehu (2004). Integrated pest management in development.
Review of trends and implementation strategies. Agriculture and Rural
Development Working Paper 5. The World Bank, Washington DC.

Staver, C. and F. Guharay (2003). Building Integrated Pest Management practices in
Central America. In K. Meredia, D. Dkouo and D. Mota-Sénchez (eds) Integrated
Pest Management in the Global Arena, CABI international, Wallingford, UK.

Strauss, J ., M. Barbosa, S. Teixeira, D. Thomas and R. Gomez (1991). Role of education
and extension in the adoption of technology: a study of upland rice and soybean
farmers in central-west Brazil. Agricultural Economics 5: 341-359.

Swinton, S. (2005).Assessing economic impacts of natural resource management using
economic surplus. In B. Shiferaw, H. Freeman and S. Swinton (eds), Natural
Resource Management in Agriculture: Methods for Assessing Economic and
Environmental Impacts. CABI Publishing pp 155-174, Wallingford, UK.

71

USDA (United States Department of Agriculture) (1998). Acute toxicity of pesticides.
Agricultural Marketing service. Pesticide Data Program. URL:www.ecologic-
ipm.con1/PDP/Table_21998.pdf.

Walker, T. and C. Crissman (1996). Case Studies of the Economic Impact of CIP-related
Technology. Lima, Peru. 157pp.

Wooldridge, J. (1997). On two stage least squares estimation of the average treatment
effect in random coefficient models. Economic Letters 56:129-133.

Wooldridge, J. (2002a). Econometric Analysis of Cross Section and Panel Data. MIT
Press. Cambridge, MA.

Wooldridge, J. (2002b). “Unobserved heterogeneity and estimation of average partial
effects.” Working paper. Department of Economics. Michigan State University.

Wooldridge, J. (2003). “Cluster sample methods in applied econometrics.” Working
paper. Department of Economics, Michigan State University.

72

MULTI-INSTITUITIONAL IMPLMENTATION OF FARMER FIELD SCHOOLS

AMONG NICARAGUAN BEAN GROWERS. DO DIFFERENT NGOS

PERFORM DIFERENTLY?

ESSAY THREE

73

1. Introduction

The reduction of the public agricultural extension services in many developing
countries has induced the entrance of new extension providers. Among these new
providers, non-govemmental organizations (N GOs) have received special support from
international donor agencies (World Bank 1991, Wallace 1997) and have increased their
participation in the delivery of agricultural technologies. Oﬁen, NGOs are seen as more
efﬁcient and cost-effective extension providers than governmental entities (F arrington et
al 1993, Edwards & Hulme 1996, Reuchelin 2003) and as a better means to reaching poor
farmers (Carney 1998). The increasing participation of NGOs in extension systems has
also increased the pluralism of providers, usually highlighted as a desirable condition for
increasing the extension supply (Qamar 2002, Sulaiman & Hall 2004). However, many
other papers have warned that pluralism also introduces a variety of organizational
characteristics and could affect the extension performance (Rivera & Gustafson 1991,
Hashemi 1992, Hassan 1993, Robinson 1993, Christopolos 1996, Garforth & Lawrence
1997)

The impact of agricultural extension services has been broadly studied, but most
of these studies have concentrated on measuring the impact on farmers’ knowledge,
technology adoption or farm productivity (F eder et al 1987 Birkhaeusen et a1 1991,
Strauss et al 1991, Hussain et al 1994, Owens et al 2003, Akobundu et al 2004). It has
been recognized that impact assessment usually ignores the institutional framework of the

extension process and the characteristics of the actors who facilitate this process (Raina

74

2003). The issue of how an increasing diversity of NGOs engaging in extension activities
affects extension outputs has received no attention in the literature.

This essay analyzes how different characteristics of NGOs working as extension
providers affect the delivery of agricultural technologies to farmers beneﬁciaries of this
extension service. To achieve this objective we analyze the Nicaraguan experience
implementing a multi-institutional project of farmer ﬁeld schools (FFS) for disseminating
integrated pest management (IPM) among bean growers. Using farm level data this essay
1) measures the effect of FFS participation on farmers’ pesticide use, adoption of IPM
practices and bean net beneﬁts and 2) evaluates whether the FPS effect is inﬂuenced by
the different characteristics of the different NGOs participating on the FFS
implementation.

This paper ﬁrst provides a summary of the increasing participation of NGO in the
extension services of developing countries in section II and speciﬁcally describes the
Nicaraguan case in section 111. Section IV presents the analytical framework of the paper
that includes the sample design and the empirical strategy. Section V presents the results
and ﬁnally section VI summarizes policy implications and conclusions.

2. NGOs and their participation as extension providers in developing countries

Public agricultural extension in developing countries has been largely criticized
for not being relevant and for having produced insufﬁcient impact (Rivera & Gustafson
1991, Chapman & Tripp 2003). In addition, the recent wave of structural adjustment
programs has produced severe budget cuts that have reduced the presence of
governmental entities in the national extension services (Farrington 1994). The generated

shortage of extension services in developing countries induced the entrance of new

75

providers that include non-govemmental organization and private institutions (Carney
1998, Qamar 2002).

The debate persists on whether developing countries can succeed in promoting
broader private extension services (Rivera 2001, Eaton & Shepard 2001, Rivera & Zijp
2002). So far many papers have highlighted the private sector’s limitations for reaching
all potential beneﬁciaries (Berdegué 1997, Rivera 2001, World Bank 2002). Non-
governmental organizations are considered more efﬁcient and cost-effective providers
than public institutions (F arrington et al 1993, Edwards & Hulme 1996, Reuchelin 2003)
and especially the best mean to reach the poorest and to develop strong relationships with
local farmer organizations (Carney 1998, Farrington and Bebbington 1994). These
characteristics have led international donor agencies to support greater participation of
NGOS in developing countries’ extension systems (World Bank 1991, Wallace 1997).

NGOs have been encouraged to initiate extension services because a pluralism of
extension providers is considered desirable for the new conceptualization of national
extension systems (Qamar 2002, Sulaiman & Hall 2004). However, this pluralism has
also introduced a diversity of institutional characteristics that vary from institution to
institution and have led some studies to question the accepted wisdoms of NGO
advantages for delivering agricultural technologies (Bebbington & Thiele 1993, Vivian
1994). Some NGO characteristics favor successful extension outcomes but others
produce an ambiguous effect on extension performance or to even impede it. Larger
NGOS are expected to perform better because they have more resources and extension
agents to deliver technologies (Hassan 1993). However, smaller NGOS tend to have

better local knowledge and a better relationship with farmer communities that are

76

necessary for a successful extension 'program (Farrington & Bebbington 1994, Garforth
& Lawrence 1997). NGOS are more likely to succeed when they manage few and simple
extension projects (Christopolos 1996). However, for the sake of increased funding tend
to make NGOs sometimes accept a large number of projects, thereby decreasing the
efﬁcacy of each of them (Edwards & Hulme 1996). Funding pressure has also forced
NGOs to expand their geographical areas of inﬂuence, but often at the cost of not
reaching the poorest (Hashami 1992) and expanding into activities less relevant for
national development (Robinson 1993). Many NGOs have also been criticized for having
staff with inadequate scientiﬁc training (Garforth & Lawrence 1997, Chapman & Tripp
2003). Longer experience in speciﬁc areas can constitute an advantage for NGO’s
disseminating technologies, but new NGOs have an unavoidable trend to start their
extension service only in new areas (Carney 1998). Finally, a broad number of NGOS
participating in the same extension project can also introduce diversity among
institutional interests about the emphasis that the project should follow (i.e, priorities on
project activities, targeted beneﬁciaries, etc) affecting the extension performance. Many
NGOs tend to emphasize activities of interest of their main funding providers that do not
necessarily match with speciﬁc project goals and (Van der Ban 2000, Rivera & Qamar
2003)

Multi-institutional extension projects could beneﬁt from a pluralistic institutional
environment, but the diversity among each of the institutions participating in these
projects can also condition the performance of the extension outputs. This empirical
question has however been absent on previous impact studies about delivery of

agricultural technologies. This paper takes a ﬁrst step toward inclusion of these

77

institutional characteristics in the analysis of agricultural extension impacts and tests
whether extension provider characteristics related to size and resource capacity, expertise
and experience, and institutional focus affect farm outcomes of participants in extension
programs. The rest of the paper analyzes a case study for implementing a recent multi-
institutional project of farmer Field Schools (FFS) in Nicaragua.
3. The extension service in Nicaragua and the delivery of IPM through a multi-
institutional FFS project

Agricultural extension started in Nicaragua in the 1970’s with the large public
extension structure developed with the support of the US. Agency for International
Development (U SAID) and was maintained during the 1980’s by the Sandinista
government. In both periods, extension services were targeted mainly toward large
farmers (Christopolos 2001). At the beginning of the 1990’s a ﬁnancial crisis in
Nicaragua led to reform of the public extension service, resulting in the creation of the
Instituto Nicaragtlense de Tecnologia Agropecuaria (INTA) and the entrance of the ﬁrst
NGOs into the extension system. The major reform introduced was the expansion of
extension services to small farmers (Christopolos 2001). The number of extension
providers remained relatively low up to 1998, when after Hurricane Mitch, a wave of
international funding promoted a massive entrance of more NGO’s with a diversity of
characteristics to the extension service (Lavard & Marin 2000). In 2001, with World
Bank support, the national government started a major reform of public extension
programs reducing the presence of INTA in many areas and promoting the creation of
more NGOs to replace INTA work in some areas (Barandun 2001). The inﬂux of new

extension providers occurred at a time when outreach programs were increasingly called

78

upon to difﬁrse complex technologies like integrated pest management (IPM) (Bentley &
Andrews 1996, Hruska and Corriols 2002, Staver & Guharay 2003).

Integrated Pest Management (IPM) is a group of pest control methods aimed to
reduce environmental and health risks to farmers by keeping a crop pest infestation below
an economic threshold level. This level is the pest population density at which control
measures are necessary to prevent a decline in net returns (Femandez-Cornejo et al 1998,
Bajwa and Kogan 2002). Pest control methods may include pesticides when necessary
but also non-chemical inputs and specialized practices such as agro-ecological analysis”,
botanical insecticides, insect sticky traps and others (Femandez-Cornejo et al 1998). In
Central America the ﬁrst three IPM practices has been widely disseminated as non-
chemical pest control and are the focus in this essay.

IPM extension in Nicaragua has mainly followed the training and visit (T&V)
approach (Benor et al 1984). The T & V system is based on short ﬁeld visits to selected
farmers who are put in charge of delivering technical packages to neighbor farmers
(Picciotto & Anderson 1997). It has been criticized for being “top down” and for failing
to organize farmers (Hussain et al 1994).This traditional extension approach has usually
been indicated as the main responsible for the low rate of IPM adoption in Nicaragua
(PROMIPAC 2001).

In order to improve the adoption of IPM, the Project for Integrated Pest
Management in Central America (PROMIPAC) has promoted the implementation of
Farmer ﬁeld Schools (FFS) since 2001. FF S were ﬁrst established in Asia at the end of

the 1980’s with the intent of increasing the dissemination of knowledge among farmers.

 

'3 This practice consists in a permanent observation of the crop ﬁeld and the use of insect scouting as the
main activity. It will determine whether pest infestation has reached the economic threshold level.

79

FF S uses participatory techniques and combines the scientiﬁc knowledge and the practice
of IPM with farmers’ experience and interests under the learning by doing approach
(Gallagher 1998). With this new approach, farmers now incorporate their priorities,
develop more skills, and become the main actors in the decision making process of
technology adoption (Kenmore 2002). It is also expected that FF S graduates will be more
able to convince other farmers about the value of IPM and thereby accelerate its adoption
(Fleischer et a1 1999).

Following the existing trends of extension services in many developing countries,
FFS in Nicaragua have been implemented by a group of NGOs trough a multi-
institutional project. As stated before, this extension approach increases the number of
extension providers available for targeted farmers, but it can also introduce a diversity of
institutional characteristics related to extension capacity and focus of each NGO that may
condition the delivery of IPM. Some of the participating NGOs had had experience in
delivering IPM, but the others had their ﬁrst IPM experience with F F S. Most of the
NGOs with no previous IPM training experience grew out of the partial privatization of
the extension services funded by the World Bank.

Differences among NGOs participating in the F FS project are not restricted to
past experience with IPM. They also include differences in NGO size, resources for
delivering IPM and institutional interests. As shown in Table 3.1, the NGOs participating
in the F F S project differed in number of total extension agents, extension agents trained
in IPM or in FFS, number of projects being developed by each institution and the area of
inﬂuence of each of them. Also the seven NGOs present different institutional emphasis

in their extension work that range from credit programs to soil conservation practices

80

.EoEtagoQ 28 m0 350 06 9 vacuum“: 8a 8593 :5 .308 89.2 E x83 mowZ 39:. a.

 

 

 

aouaﬁcmwco «903.8 839280 an
o :82 .565 m N m w O<ZD 88:333.. up 3:282 :35
£03 883 390E252
n .cosmtomaoo =om m _ N o m<mmmo 23589 3 8mm :Bomnaawco
30.8 .5365 3.8% o=obamoD _o a
o 0:890 £35 m o N m MmDE :Eomwumoia 2 E8 aﬂomvasm
mogsooQEmaa
3 8&8 £33m N N N w <Umhmmm 825.8. momomtom ow Begum
Eugenia/co 3E0; _o 83 0:05me
N 1.83 5803 ﬂ m m N 0305085 bow on 863mm on 8:80
5:3ng
v :82 4620 w v o o m<HE<U *Ammouoﬁnv «5385 3:80
593:wa :32 128800 «Baum/x 23.580 x
.4 .coﬂatomcoo mom 0 N m v U<<Q< 568$?on Ema comummoomaa
wcmﬁab wEEg
3253?: "doomed mum 37:
305mg £32,380 mo 55» 55’ 38,—. 8,3834 5:33:—
.«o “3:52 EcowBEE “38:2 35%“ 229888 CO 39:82

 

«$9332 5 53558295 mhh :e wan—.83 .562 he 553.889 .—.m 03:.

81

(Table 3.1). These differences could have affected farmer participation in IPM training or
their subsequent likelihood of adopting IPM.

The starting point for FF S in Nicaragua was common to all NGOs: Each of them
sent some extension agents to participate in a two-month intensive training-for-trainers
program. This training program was conducted by F F S trainers experienced in
participatory research methods. At the end of the training experience, extension agents
were expected to understand their new role as facilitators, replacing the T&V role of
delivering agricultural technologies (PROMIPAC 2001). With the variety of participatory
techniques that facilitators learned in the training-for-trainers’ course they were also
expected to be able to offer different alternatives to solve farmers problems, especially
those related to pest control. Differences among individual FFS curricula conducted by
different NGO’s should thus only be attributed to different farmers’ preferences.

The implementation of FFS in Nicaragua, however, brought some differences
related to the special emphasis that each NGO decided to give to each FF S under its
control. Table 3.2 shows the individual curricula developed by each of the 13 FFS for
bean producers in Nicaragua. Except for the fact that all bean FFS promoted agro-
ecosystem analysis, and almost all of them used botanical insecticides”, each curriculum
was developed differently according to individual NGO priorities. Only CECOTROPIC,
FIDER and UNAG promoted yellow sticky traps in their FFS, because only these

institutions had experience with this practice. ADDAC and ODESAR were the only

 

'4 The use of botanical insecticides was however very diverse among FF S according to individual NGO
experience and knowledge. For example, FIDER promoted the use of detergent + oil and chile (Capsium
anum) + alcohol, CARITAS promoted the use of chile + garlic + onions and Neem (Azadarichata indica),
UNAG promoted the use of madero negro (Gliricidia cepium), sugar and ashes, while ESETECA did not
promote any.

82

 

 

 

X X X 05,5 203855 .m

X X X 05/5 3:25 84

X X X O<ZD 23,—. .m

X X X O<75 ~ moan:

X X X ﬁmmno «ES a. 8.50

202 202 8:033 X X X amen :32. _m

222 202 833:3 X X meE 28525 _m

$3 $3 :2§=:£ x x x amen magma

$3 202 coumﬁztom X X <UMHMmm 25$”—

mmoq 222 X <Umhmmm 322:0 93

$3 $3 mocotm> X X X UEOMHOUmU 323. 85m

$3 33 X X m<HE<U BE 880

$3 304 X X X X U<<D< N 85:
35>»: .Eomcoo am: 82383:— mam—«5‘

.02 20; ~28: =om Bo=o> 32.5.5 4988.238. £58800
3:035:80 .m> En: acoEtomxo “comm—26v 5:355 mm”—

mo ”Samba—=8 20E mum 85388 350 2:355 mum E vows—2: 3:332“ 2x:

 

saw—«£2 3 m5— :3: :23 .3 123.95.. 85.52. .35: was «1.3.-.50 .N.m 03:.

83

NGOS that included soil conservation activities in their FF S curricula. In these cases,
both NGOS decided to solve first the problems of soil erosion and low productivity that
they identiﬁed as their work priority in the communities where they implemented FF S.
Although health effects derived from pesticide use should be a special component of IPM
training, only 3 FFS included explicitly this activity in their curricula (Table 3.2).

According to the training received by facilitators, ﬁeld experimentation should be
a strong component in each FF S, and the comparison of a plot under IPM management
with a plot under a community traditional management was suggested as part of each F FS
curricula. However, ODESAR and UNAG did not conduct this experiment (Table 3.2).
Also not all the FFS conducted other complementary experiments (only 6 out of 13) and
the type of experiments were usually different across FF S (Table 3.2).

Differences among institutions in charge of the F FS project and differences in the
implementation of individual FFS could have conditioned the overall FFS impact in
Nicaragua. This paper examines this empirical question by testing whether individual
characteristics of NGO serving bean growers enhanced or limited the F FS impact on F F S
graduates’ adoption of IPM practices, pesticide use and bean net revenue (whether these
NGO characteristics constitute part of the FF S treatment effect).

4. The Analytical Framework

The evaluation of program impacts is usually done using the counterfactual
analysis where targeted outcomes are measured for some individuals receiving the
program (treated group) and for some individuals that do not (counterfactual group)
(Rubin 1974, Ravallion 2005). This paper extends this methodology for evaluating the

average impact of IPM training on farmers by examining the effects of individual NGO

84

Characteristics. Given the multi-institutional nature of the FPS implementation in
Nicaragua, this paper incorporates a group of NGO characteristics that differs among
participating NGOS that could enhance or limit the FFS impact on participating farmers’
outputs. Characteristics related to NGOs’ size and resource capacity, NGOs’ expertise
and experience with IPM and targeted farmers, and NGOs’ institutional focus are
included in this program evaluation framework.

Under an agricultural household model framework, we assume that an FFS
educational effect is linked to household utility through improved knowledge (F eder et al
2003). Such improved knowledge about pests and pest controls could potentially
inﬂuence farmers’ input decisions and farm net revenues. We propose to test
econometrically for this effect on input use, as well as to test the hypothesis that the
delivery of knowledge could be inﬂuenced by individual characteristics of NGOS in
charge of the delivery programs. In the rest of this section we provide details of the data
collection and econometric strategies for conducting this program evaluation.

4.1. The sample design and data collection

A set of farm level data was collected between May and August 2004 with a
cross-sectional survey of 436 households of Nicaraguan bean growers in the departments
of Esteli, Matagalpa, J inotega, Madriz, Masaya and Carazo. The sampling design
followed a double stratification (Deaton, 1997) to compare the effect of different IPM
training methods (FF S vs. T&V) and to include diverse settings. Households were
interviewed in 74 rural communities, including 13 where FF S were implemented, 9

where FFS graduates lived but no FFS were held, 26 communities selected randomly

85

where no FFS exists but other IPM extension services were available”, and 26
communities selected randomly where no IPM extension was present. In each
community, households were selected randomly and included clients and non clients of
NGOS. The sample distribution includes FFS graduates, farmers participating in other
IPM programs, FFS graduates who also attended other IPM programs, and farmers who
no prior contact with formal IPM extension.

4.2. The econometric estimation

4.2.1. Potential econometric problems

The main potential econometric problem in this paper is the endogeneity
associated with self selection of farmer participation in IPM extension programs and the
non-random placement of these programs. Many studies have recognized that failure to
correct this problem produces biased estimates of the extension impact, which usually
result in an overestimation of their impact (Strauss et al 1991, Owens et al 2003, F eder et
al 2003).

In Nicaragua all NGOS participating in the F FS project agreed to implement a
group of general rules while organizing each FF S. This group of rules included selecting
a place for meetings and experimentation with good access, relying on existing farmers’
organizations and farmers’ willingness to participate on IPM training. However, most of
these rules were related to pre-existing farmer-NGO relationships, which were usually
not observed during the data collection process. Farmers with good pre-existing
relationships with NGOS could have been more willing to participate in IPM training,

whereas farmers with poor NGO relationships could have been less so. It is clear that FFS

 

'5 Other IPM extension services refers to the training and visit (T&V) approach that was mainly provided
by the same NGOS in charge of FF S implementation but usually targeted for NGOS clients who did not
participate in FFS.

86

program availability was not randomly distributed among Nicaraguan farmers and that
selectivity bias could be present in the econometric estimations.

This paper evaluates ﬁrst the convenience of using two stage least square (2 SLS)
to correct selectivity bias in the econometric estimation, which is the preferred method in
linear models with cross sectional data (Wooldridge 2002a). The model for the adoption
of IPM practices has a clear non-linear nature and would be more suitable for a trivariate
probit or multinomial logit speciﬁcation, but correcting selectivity bias in non-linear
models through the recommended “control function approach”l6 generates big
computational problems and can only be applied under normality conditions (Wooldridge
2002a, Wooldridge 2002b, Blundell & Powell 2004). But as Angrist (2001) pointed out,
standard instrumental variable estimation can still be appliedto non-linear models when
the objective is to estimate the causal effect of the endogenous variables. Wooldridge
(1997) also showed that 2SLS could still be applied in a random coefﬁcient model with a
binary endogenous variable when this variable follows a linear probability model, the
linear approximation of models with binary dependent variables as the adoption of IPM
practices case. Even recognizing the non-linear nature of the IPM adoption model, this
paper will maintain 2SLS as the preferred method to control endogeneity of IPM training
participation.

The sample design poses secondary econometric problems of a clustered and
stratiﬁed sample that can bias the parameter estimates (Wooldridge 2002a). To deal with
the clustering problem, this paper adjusts the variance matrix and includes dummy

variables for each cluster using survey regression methods (Wooldridge 2003). This

 

'6 The control function approach adds to the estimation equation an additional variable that controls for part
of the correlation between an endogenous variable and the error term (Petrin & Train 2005, Blundell &
Powell 2004).

87

method is suitable given the nature of the sample design that has small clusters (six
households per village on average) and a large number of clusters (74 villages). To
correct the unbalanced representation of farmers (especially F FS trainees) in the sample,
this paper uses a weighting scheme as suggested by Wooldridge (2002a) (see appendix
A3 for details).
4.2.2. The econometric speciﬁcation

This paper speciﬁes three linear models for pesticide demand, one linear model
and three linear probability models for the adoption of IPM practices, and one linear
model for the bean net revenue function. All models have the same set of explanatory

variables and the general model structure is:

Xj =1} 1531f”; :Bjr +IN 'le +E‘FSIN :BJ'FFSN +CH ﬂjCH+CC ﬂjCC+ljj ;Forj=1,...,8

The jth dependent variable depends on vectors of K output and input prices (PK),
household participation in the T IPM training program (T1), the N individual
characteristics of the NGO delivering IPM to the household (IN), the interaction of FFS
participation and NGO characteristics (TFFsIN), as well as socioeconomic and other
household characteristics (CH) and community ﬁxed effects (Cc), with disturbances
assumed to be independently distributed (U). All variables used in the econometric
estimation are described in detail in the following subsections.

This paper uses the predicted probability of each IPM training program
participation (FFS and other programs) as generated instruments of the true program

participation in all the 2SLS models. Each predicted probability is estimated using a

88

probit speciﬁcation that has as explanatory variables the original exogenous variables
plus some redundant variables that explain the variation of FFS and other IPM program
participation (Wooldridge 2002a). The structure of each probit speciﬁcation uses the
same set of explanatory variables as the previous models plus a vector of Z redundant
variables that are related to participation in IPM training, but believed to be uncorrelated
with the disturbance terms of the IPM adoption, pesticide demand and bean net revenue
equations (see ﬁrst table of Appendix A.9).
4.2.3. The dependent variables

Household input demand for insecticides, herbicides and fungicides is represented
by the quantity of pesticide active ingredients used by each household in bean production
during the most recent season in 2003 weighted by acute human toxicity. A human
toxicity index is calculated for each household i as the sum over all m pesticide active
ingredients (aiik in kg/ha, k=1,. . .,m) divided by each active ingredient’s mammalian
toxicity (measured by the minimum dose per gram of body weight that is lethal to 50%
of a test rat population or LD50, as reported in USDA, 1998). The human toxicity index,

show below, is proportional to the LDSO and is increasing in lethality.

_ all):
HT!" _ 2* LDS 0,,

 

The adoption of the three speciﬁc IPM practices promoted in F FS is represented
by binary variables that indicate whether a household adopted agro-ecological analysis,

botanical insecticides and/or yellow insect traps and a variable that measures whether a

89

household adopted one, two or the three IPM practices. As explained before this paper
uses a linear approximation for these non-linear models.

Bean net revenues is a continuous expected variable measured in US$/ha and
calculated based on farmer self-reported revenues and costs during the most recent bean
season in 2003 prior to the household survey in early 2004.

4.3.4. Explanatory variables

Output prices in the regression models are selling prices for beans and maize reported in
US$/kg. Input prices include bean seed, metamidophos insecticide, gramoxone (paraquat)
herbicide and mancozeb fungicide”. The bean net revenue function also included the
price of fertilizers”. Other prices and price proxies are the transport cost to the nearest
market of a sack of beans or maize and the distance between each farm and the municipal
capital. Other production variables include the most recent farming season that bean were
grown: primera, postrera or apante, the use of disease resistant bean varieties, altitude
(proxy for rainfall and temperature) and farmers’ observed level of bean pests and
diseases (high or not).

IPM training participation is speciﬁed for F FS graduates, other IPM training
program participants and the interaction of the two”. Households without IPM training
contact are kept as the control group. Some characteristics of the participating NGOS that
usually differ among them are explicitly speciﬁed. We use variables that measure NGO
size and resource capacity (average number of extensionists per NGO project, average

number of NGO extensionists per district), NGO expertise and experience (proportion of

 

'7 Metamidophos, paraquat and mancozeb were by far the main active ingredients among insecticides,
herbicides and fungicides on the Nicaraguan farms surveyed in 2004

" we use the price of the most common combination of nitrogen, phosphorus and potassium (12-30-10)
'9 There were 35 FFS graduates who had previous participation in other IPM training programs

90

NGO staff with IPM and FF S training, and years of experience of NGO working in the
respondent’s community), and NGO institutional focus (whether the NGO has a main
focus on soil conservation , or agricultural credit, whether the NGO conducted IPM
experiments through FFS, Whether farmers observed greater yields in the experimental
IPM plot, whether farmers observed greater net revenues in the experimental IPM plot
and whether the NGO organized other complementary experiments during FFS
implementation). Interactions for FF S participation and individual NGO characteristics
were also included as explanatory variables in order to measure whether each individual
NGO characteristic affects FF S impactszo.

Household characteristics are represented by two groups of variables: (1)
measures of household assets: farm size, area under irrigation, number of cattle, whether
family has electricity at home, and whether the household receives remittances from
relatives working in foreign countries; and (2) measures of household composition:
gender of household head, age and years of education of household head, percentage of
females, and percentage of family members under 14 years old. Other conditioning
variables include whether the household relied upon hired labor or family labor for
pesticide application and whether household members used protective devices like mask,
gloves or special clothing while applying chemicals. Binary (dummy) variables are
included for community-level ﬁxed effects with one of them kept as a control. The 73
community dummies are used for the ZSLS estimations, but are substituted by
municipality dummy variables (9 out of the 10 municipalities) in the estimation of the

predicted probability of FFS and other IPM training participation.

 

2° Therefore interaction for the predicted probability of IPM training participation (instruments) and
individual NGO characteristics were also generated for all ZSLS estimations.

91

Finally, redundant variables are used to control for endogeneity of extension
participation. All these variables measure pre-existing linkages between individual NGOs
and client farmers, including whether households received previous credits from the
NGO, whether they received food assistance or any cash support, and whether they
adopted soil conservation practices prior to their IPM training (terraces, contour lines and
live fences). As showed in Appendix A.9 many of these redundant variables meet the ﬁrst '
condition for a good instrument and signiﬁcantly explain farmer participation in FFS and
other training programs. These variables also meet the second condition for a good
instrument in that they were believed to be uncorrelated with other variables that explain
the variability of the dependent variables of the household demand models (disturbance
terms) but not observed by the researcher. Credits and other NGO support received by
farmers prior to the most recent bean season should not inﬂuence farmers’ decisions
about the level of pesticide use or the adoption of IPM practices during this recent bean
season. Likewise, prior adoption of soil conservation practices should not condition
farmers’ decisions about pesticides or non-chemical pest management practices used
during the most recent bean season. The adoption of terraces, contour lines and live
fences are not related to the level of pest and diseases and therefore the related pest
management practices used. The characteristics of these variables make them suitable
for specifying robust IV models that will be tested for correcting endogeneity of farmer
participation in FPS and other IPM programs.

5. Results
As stated before, F F S in Nicaragua were carried out by NGOs whose diverse

institutional characteristics (Table 3.1) generated heterogeneous implementation

92

approaches (Table 3.2). In order to estimate the effect of this institutional and individual
FFS diversity on farm outcomes, the econometric analysis uses the institutional variables
listed in Table 3.3. It starts by performing Hausman tests in order to evaluate the need for
endogeneity correction using 2SLS models (Wooldridge 2002a). As shown in Table 3.4,

Table 3.3. Descriptive statistics of selected variables used in the survey least squares
for 436 Nicaraggan bean growers, 2003-04+

 

 

 

Variable description Average Std dev
Dependent Variables

Index of insecticides weighted by toxicity (x 100) 2.9 5.1
Index of herbicides weighted by toxicity (x 100) 0.5 0.7
Index of fungicides weighted by toxicity (x 100) 0.1 0.6
Adopted agro-ecological analysis (%) 28.2

Adopted botanical insecticides (%) 35.1

Adopted insect yellow traps (%) 15.8

Adopted only one IPM practice (%) 19.1

Adopted two IPM practices (%) 14.9

Adopted three 1PM practices (%) 10.8

Household net revenue 91.2 124.9
IPM training and NGO characteristics

FFS participation (%) 20.6

Other IPM training participation (%) 34.2

FFS and other IPM participation (%) 8.0

Number of extension agents / project 1.5 1.4
Number of extension agents / district 1.0 0.8
NGO staff with IPM training (%) 45.9

NGO staff with FFS training (%) 27.9

Years of experience NGO 8.1 6.6
NGO emphasis in soil conservation (%) 7.1

NGO emphasis in credit programs (%) 10.3

Performed comparative trials (%) 36.0

Performed complementary trials (%) 28.2

IPM plot greater yields (%) 15.6

IPM plot greater net revenues (%) 11.5

 

+ Other variables used in econometric analysis are shown in Table A. l .3 (appendix)

the hypothesis that IPM training participation was exogenous is rejected in all models,
suggesting that the IV estimates of the pesticide toxicity models (insecticides, herbicides
and fungicides), the adoption of IPM models (agro-ecological analysis, botanical

insecticides and insect traps), and the net revenue model are more suitable.

93

Controlling for the endogeneity of F FS participation reveals several OLS results
that misleadingly appear to show F F S impacts when in fact they are due to nonrandom
selection of participants. FFS participation by itself had no signiﬁcant impact on the
adoption of IPM practices, pesticide use or bean net revenues (Tables 3.5, 3.6 & 3.7).
These ﬁndings are consistent with similar studies in Indonesia (Feder et al 2004).

Table 3.4. Hausman test and joint significance F-test of specific NGO characteristics
interacting with the FFS treatment effect on household pesticide demand, adoption
of IPM practices and bean net revenue models for 436 Nicaraguan bean growers,

2003-04.
Hausman test FF S aggregate effect

 

F -Statistic P-value F-Statistic P-value

 

Pesticide demand

Insecticides 2.05 0.0152 2.42 0.0034
Herbicides 1 .80 0.0388 ‘ 1.90 0.0267
Fungicides 3.65 0.0000 0.92 0.5460
IPM adoption

Number of IPM practices 3.01 0.0003 5.54 0.0000
Agro-ecosystem Analysis 6.69 0.0000 3.66 0.0000
Botanical Insecticides 1 .61 0.0755 4.1 8 0.0000
Yellow insect trap 5.54 0.0000 4.28 0.0000
Bean net revenue 2.53 0.0022 2.22 0.0077

 

The participation on other non-F F S IPM training programs performed slightly better in
Nicaragua, increasing the likelihood of adopting botanical insecticides by 60% and insect
traps by 52% (Table 3.5). Some of these ﬁndings can be explained by poor results on the
FFS demonstration plots. FFS graduates could observe yield gains from IPM
experimental plots in only four of the 12 FFS and net revenue advantages in only two.

Most of the FFS graduates observed superior yields and net revenues on the bean plot

94

 

 

95

.88 .28 .38 .88 .8. 8 3.8 an. 8 5.8
8.... 9...... $2. 3.... 3... 9...... 85 as... .22. .8. 8:88
.88 .8. 8 .2. 8 .2. 8 .8. 8 .38 .38 .58
3.2399: ©_a.o «n—Mé ~wod cavué actmnmé acnd «ca—N04 83:26.— 3: 0.0:.— 83300
8...... .28 a... 8 .28 .88 .38 .2... .38
5...: 82 .3... .2... 8...... .:n.....m .23..." 2:38... .20.. 2o... 85......
.88 .28 .88 3.8 .2. 8 8.8 2...... .88
mﬁcé ﬁsamhé mend mood .52..— «cawbmé — —c.¢ «.535.— mEoEmbn—xo .550
.3. 8 .58 8...... .88 .88 .88 .88 G... 8
:E- 2n... 3.... .2... :3. 2:2..- Rn... 8... .- €05.85 38.5.58
.8. 8 5.8 3.8 6...... 8...... .88 .8. 8 a... 8
8%..- 8.... N8... .2... SN... :82... 3...... 8m»... 9582.. .82.. a. 88.3.”.
.88 .88 .88 .88 .....8 .88 .88 an...
3...... 28... .8... :2... 2.... 8.2.. $3 5.... 5.8528 =8 5 8.23m
.38 .38 .38 8...... .38 $2.. .8. 8 .28
8...... 8m... 3...... .8... 8.2.. 2”... 2m. .- 8...... 8.5.... ma. 5.3 8.5.83...
.88 8.8 .8. 8 8... 8 G... 8 .88 2...... .88
2...... 83 .89... a...“ 1.8... 5.8.3. 8...: 5.8...." was... 2... 5.3 8.8.28...
.38 9...... an... 5.8 8.2.. .3... .28 .8...
8...... 2...... 8...? .8... ”8.... 2...... 2...... 2...... 85.3....» .o as. 002
.38 8...... .38 .8. 8 8...... .88 .88 .88
N3... :2- 8...... N3..- 83.- :83- :3. :32 .- are... .2. 8.8.22...
38 .38 .28 .88 8...... .28 8...... 8.8
22- 8.... 8:. 2n. 3.... :32...” .8... :5»... .88... .2. 328.838
mm... 5.: 22.8.3...
8.8 .88 G... 8 82 an... an. 8 .3. 8 a... 8
2.... 8...... Ex... :38... 8.8 «8.... 8.2.. :88... mass. 2... 2.8.5
.28 can. .28 an... .8. 8 .38 .88 5....
cab—m6 «earn—é «ccbmmé «ccvnvé Oo—d cacaQ—é tin—N.— c¢chﬁh6 wcmﬁg 2a: .550
.38 G... 8 .88 a... 8 .28 .38 .28 8&8
82.. .82. 8:... E... 82.. :23. .8. .- :38... m“...
38. >_ an... £88.53 2 £88.82. warsoum >_ Iagoom 8088.8 .2 >_ 80..qu .Z

 

+2733 6.83....» 5.2. namwgaﬂz
emv 3:50.. 8339.»...— zotsn 8.3.5... «muon— doozuﬁa 2.: no newt—88a E. 390...... 833.8322? OGZ 8.... mm..— m.m 03.5.

 

88:28 323.8923 mo .0388 £2: 05 m. 95:08. 83.5 I

8;... 585.35 E 8327. 2.. 8:52 8282?. 82.5.00 +
885:an E 85.3% . .«o 33> 8283..

a... a 2.88:»; z... 2% a 2.85%.: : 8... a 28.835 ._

 

8888.8
end

3.8
one

.28
2....

3.8

3...
.38
Sm...

mmhcd
5..

mm...
89.

.38
8a...

and

one
.38
:32:.—

8888.8
$6

8m...
82.

.28
.8...

mm...

.8.
$8
3......

888.8
.86.

cm...
09.

.28
3......

3.8

09
AQNV
:38."

33>-..
08.0.8.3-..
.8. 5.5.8:
825:3...—

mcomuutowno

2.5280

96

 

 

.38 .NN.8 .3 8 .8. 8 .K. 8 and.

8...... 8...? :8... 2...... .2... 2:3... 2...... $2 Baa...

.28 .38 .8. 8 .38 .38 .88

8...... :88..- .33.- 8...... 2.3.... 5mg..- 32.2.2 .2. 22.. 828....

.38 .. ..8 .38 .2. 8 2...... .88

3.....- 3...? 3...... 3.....- 8.... 8...... .20.. 22: 3:88

.88 .3. 8 .ﬁ. 8 .28 .38 .38

8...... .88..- £8..- 8...? 3.... N8..- 355:2...» .2...

.38 an... .88 5.8 .28 .38

3.... 8.... 8.... 8...... : .... n8... 955:2...» 2.53.58

.88 .2... .3... . .38 .28 .38

N8... 8...... h 8.... 5.....- 8.... 8.... 25.2.. .86 5 .888...

8...... an... .88 .38 .38 .38

8...... 8...... .8... :38..- at... 8...... 8.3588 =8 5 gain...

.88 .3... .38 .S. 8 .3. 8 .28

8...... 8...... 8...... 8...... 3...... 2...... .558. m... 5.3 9.8.82...

.28 .88 .3. 8 .E. 8 3.8 .38

.8... .. s... :2”..- .28...- 8.... 8:... .582. 2... 5.3 9.8.22...

.88 .58 .38 .38 .88 .S. 8

2...... 8...... .88..- 8...... 2...... 8...... 85.593 .0 as. 082

.88 .38 .38 .2. 8 .38 .88

v8... 8.... 2.8.... .3... an... 3...... 5...... .2. 3.8.2....m

.88 .38 .38 .88 .3. 8 .28

8...... 8...? :5..- .32...- 2...... .8... 3%... .2. 3.8.23...
mm... a... 2.....23...

...~. 8 .88 .3. 8 .38 .3. 8 .3. 8

8...... :88... .25..- :25..- .32.... as..- was... 2... 28.8

.88 .28 .38 .88 .38 .S. 8

3...... 2...... 8.... 2...... 2.... 2...... was... 2... .050

6...... a... 8 an. 8 .38 .R. 8 .38

2...... .8... 2...... 32...... EN... :33... m"...

828%... >_ 028%.: 32058... >— mu.._u_€om 820.88.: >. 820882:

 

+3.83 .393...— EE: 5. «.332

one .8 338.. 3.39.»?— mot—5 83...... 3.8.. 6:35.. «39:88 E. «3&0 85232229 902 ES mhh 9m 03:.

 

97

 

o.< 58:08.3 5 :32... o... 3.32 80.30%»: 829.80 +
88553.. E 85.8% H 8.. 2:? 8385.
e... a €2.88... 2: a... a .5088»... : 3...: a 28.35 .

 

3.8

3.8

mm...

3.8

2... 9... 833...-..

 

98

employing conventional pest management (Table 3.2). Table 3.4 also includes results of
the F-test indicating rejection of the null hypothesis that speciﬁc NGO characteristics do
not alter the FF S treatment effect. We fail to reject only the null hypothesis that FFS
Table 3.7. Effects of FFS and NGO characteristics on bean net revenues. Least

squares survey regression results for 436 Nicaraggan bean growers, 2003-04+

Bean net revenues IV bean net revenues

 

 

FF S 1,001.366 -492.68
(1.24) (0.75)
Other IPM training 441.840“ -2.81
(2. 10) (0.05)
Double IPM training 320.0443 41.17
(0.44) (0.54)
Interactions with FFS
Extensionists per project 7,125.919* -233.46
(1.77) (0.29)
Extensionists per district -15,213.405* ~187.03
(1.69) (0.14)
NGO years of experience 18,280.718 359.36
(1.30) (0.15)
Extensionists with IPM training 741.935 179.43*
(0.35) (1.68)
Extensionists with FFS training -327.830*** 1.71
(2.96) (0.05)
Emphasis in soil conservation -512.6l8 469.62
(0.32) (0.75)
Emphasis in credit programs l,266.985** -178.56
(2.09) (0.62)
Comparative experiments 2,791.729* -125.28
(1.69) (0.16)
Other experiments 1,266.050 -402.67
(1.11) (1 .01)
Observed higher yields 6,913.019 1672.01
(0.97) (0.62)
Observed higher net revenues -10,161.417 1471.12
(1.36) (0.59)
Observed lower net revenues 613.107 46.49
(1.17) (0.75)
Observed lower yields -1,611.658* 160.71
(1.81) (0.59)
Observations 436 436
R-squared 0.50 0.37

 

* Signiﬁcant at 10%; " signiﬁcant at 5%; "" signiﬁcant at 1%
Absolute value of t statistics in parentheses

+ Complete regression results are shown in appendix A.9

 

99

participation and its interactions with NGO characteristics do not jointly inﬂuence the use
of fungicides.

After controlling for FF S and other IPM training participation, this paper shows
that FF S impacts on adoption of IPM practices, pesticide use and bean net revenue are
signiﬁcantly affected by individual characteristics of NGOS in charge of the FF S
implementation (Tables 3.5, 3.6 & 3.7).

The adoption of IPM practices among FF S graduates was signiﬁcantly affected by
NGO characteristics related to expertise in IPM and the institutional emphasis given to
individual F F S implementation (Table 3.5). NGOs with a greater proportion of extension
agents trained in IPM signiﬁcantly increased their FFS graduates’ likelihood of adopting
both agro-ecosystem analysis and botanical insecticides by 807% and 866% respectively.
Whether most of these extension agents were also trained in PPS methods does not have
a signiﬁcant effect on this adoption decision, which seems to indicate that training in IPM
knowledge is more relevant than training in methods for delivering this knowledge.

FFS graduates linked to NGOS that focus on credit programs were less likely to
adopt IPM practices in general and especially insect yellow traps, suggesting a deviation
from the expected IPM extension focus. They decreased the predicted probability of
adopting yellow traps as much as 100% and adopted on average 1.07 fewer IPM practices
(Table 3.5). Also farmers working with NGO’s that conducted comparative trials ﬁnding
higher revenues or higher yields in the IPM plot were more likely to adopt all three of the
IPM practices (Table 3.5) while farmers working with NGOS that conducted
complementary trials like improved seed or fertilization were more likely to adopt agro-

ecosystem analysis (scouting) by 102%. These results highlight the importance of FFS

100

ﬁeld experimentation where graduates have the opportunity to apply directly the IPM
knowledge learned. They also show that farmers observing ﬁeld advantages of IPM are
more likely to adopt it.

Pesticide use among FFS graduates was directly affected by how NGOS operated
their FFS and we measure the magnitude of the effect in terms of standard deviation of
the toxicity-weighted pesticide use. Graduates of FFS run by NGOS that implemented a
comparative trial during the FFS experimentation and observed more revenues in the IPM
plot decreased slightly the toxicity weighted use of insecticides by 0.01 s.d. and by 0.07
s.d. the use of herbicides (Table 3.6). Likewise, FFS graduates exposed to comparative
experimentation that resulted in lower bean yields in the IPM plot signiﬁcantly increased
albeit by low magnitude the use of both types of pesticides, of 0.02 and 0.03 s.d.
respectively. When NGOS decided to include other types of ﬁeld experiments like levels
of fertilization or new bean varieties, F FS graduates also reduced the use of toxicity
weighted herbicides by 0.55 s.d. (Table 3.6). Field experimentation and especially
positive results from IPM treatments seem to be highly relevant for inducing a reduction
of pesticide use. Of course, this is entirely consistent with the F FS philosophy of learning
by doing. Hence it is not surprising that failure to show FFS graduates tangible
advantages of IPM over chemical pest control can result in no incentives for changing the
level of pesticide use.

NGO capacity and expertise in IPM also affected FFS graduates’ pesticide use.
The graduates of FF S managed by NGOS with more extension agents per project reduced
considerably their toxicity weighted herbicide use by 0.31 s.d. while another year of

experience of an NGO reduced slightly the toxicity weighted of herbicide use by 0.004

101

s.d. Those from FF 8 managed by NGOS with a higher ratio of extension agents per
district actually increased the toxicity weighted use of herbicides by 0.2 s.d. In
Nicaragua, the ratio of agents per project seems to be very relevant for measuring the
time that individual extensionists can give to each group of farmers. Graduates of F FS
linked to NGOS with more extension agents trained in IPM and with more years of
experience working in farmer communities signiﬁcantly reduced the use of herbicides
(Table 3.5). This result suggests that more IPM expertise is needed among extension
providers in order to produce impacts on graduates’ pesticide use.

Finally, NGO characteristics had much less impact on bean net revenues among
FF S graduates from the 2003 season than on their adoption of pest management practices.
Only NGOS having more extension agents trained in IPM increased bean net revenues of
their FFS graduates by 179 US$ per hectare.
6. CONCLUSIONS

Impact assessment of extension services in developing countries has largely
ignored the effect of the diversity in institutional characteristics among extension
providers (Raina 2003). We found that these characteristics signiﬁcantly affect farmers’
choices of management practices and, to a lesser extent, farm net revenues. In particular,
the institutional focus, expertise in IPM and the capacity of NGOS implementing multi-
institutional extension projects signiﬁcantly affect their clients’ input decisions and
adoption of agricultural technologies.

The impact of extension programs can be enhanced or diminished by individual
characteristics of the institutions delivering agricultural technologies. NGOS with more

technical expertise and extension experience tend to enhance the delivery of these

102

agricultural technologies, as many papers have highlighted (Hassan 1993, Carney 1998).
By contrast, NGOS with an institutional emphasis different from the main extension
program deliver poor extension performance. Depending on the magnitude of these
effects, the positive effects generated by desirable NGO characteristics could be offset by
negative effects from NGOS with institutional focus irrelevant to the extension focus in
question.

The ﬁndings presented here highlight how FFS impacts can be erroneously
measured in an analysis that fails to correct for endogeneity among explanatory variables.
The uncorrected OLS estimates of FFS and other IPM training program effects are
upwardly biased, which is likely due to non-random location of the extension programs
and farmer self-selection into these training programs. So far, both of the IPM training
programs implemented in Nicaragua have had little effect on participating farmers’
pesticide use and adoption of IPM practices, two of the main goals of any IPM extension
program. However, farmers served by NGOS with a higher proportion of extension agents
trained in IPM, with greater expertise in IPM, and longer experience in working with
farmer communities tended to improve the delivery of IPM through FFS. More scientiﬁc
knowledge of IPM turned out to be more important than knowledge of new methods for
delivering IPM.

This research provides important insights for policy makers and international
donor agencies that wish to broaden the participation of NGOS as extension providers. It
is important to improve the selection of NGOs that will deliver agricultural technologies,
speciﬁcally to choose ones with strong technical capacity whose institutional emphasis

matches the main focus of the extension program. An ideal NGO in charge of delivering

103

IPM should have a large number of extensionists, the extensionists should be soundly
trained in the technologies to be delivered, the NGO should have many years of
experience working with the targeted communities and it should deﬁnitely concentrate its
extension efforts on disseminating the targeted technologies and not other institutional
activities. Our ﬁndings also underscore the major role played by FFS ﬁeld
experimentation in shaping farmer input decisions. Direct exposure to the beneﬁts and
limitations of new technologies should always be present in extension programs.
However, it is necessary to keep in mind that experimental or demonstration results may
not favor a given proposed technology. Institutions charged with improving farmer
welfare through technology diffusion should be able to react quickly and incorporate
farmers’ feedback, even to the point of discarding a speciﬁc proposed technology.

This research is a ﬁrst step toward analyzing how different institutional
characteristics among extension providers can affect the impacts of the delivery of
agricultural technologies. Future research should seek to obtain better information on the
ﬁnancial situation of participating NGOS. For example, information about amounts of
funding, relative weights of individual extension projects within the NGO’s budget, and
the amounts invested in extensionist training could expand and improve upon this
analysis. Such additional information could provide a better idea of the inﬂuence of

extension provider characteristics on beneﬁciaries’ farm outcomes.

104

References

Akobundu, E., J. Alwang, A. Essel, G. Norton and A. Tegene (2004). Does extension
work? Impacts of a program to assist limited-resource farmers in Virginia. Review
of Agricultural Economics 26(3): 361-3 72.

Angrist, J. (2001). Estimations of limited dependent variable models with dummy
endogenous regressors: Simple strategies for empirical practice. Journal of
Business and Economic Statistics 1922-16

Bajwa, W. and M. Kogan (2002). Compendium of IPM deﬁnitions (CID). What is IPM
and how is it deﬁned in worldwide Literature? IPPC publication N. 998.
Integrated Plant protection Center (IPPC). Oregon State University. Corvallis, OR
97331, USA.

Barandun, A. (2001). Developing a Mix of Public and Private Rural Extension Services:
The Experience of Nicaragua. Berater Innen News 1 URL:
www.1bl.ch/intermt/services/publ/bn/ZOO1/1/2-mix of publicpdf.

Bebbington, A and G. Thiele (1993). NGOS and the State in Latin America: Rethinking
Roles in Sustainable Agricultural Development. London, Routledge.

Benor, D., J. Harrison and M. Baxter (1984). Agricultural Extension: the Training and
Visit System. World Bank, Washington DC.

Bentley, J. and K. Andrews (1996). Through the roadblocks: IPM and Central American
smallholders. International Institute for Environment and Development.

Gatekeeper Series N. 56, London.

Berdegué, J. (1995). El sistema privatizado de extension en Chile: 17 afios de
experiencia. In: J. Berdegué and E. Ramirez (eds) Investigacio'n con Enfoque de
Sistemas en la Agricultura y el Desarrollo Rural. Santiago, Chile, RIMISP pp.
329-339.

Birkhaeuser, D., E. Robert and G. F eder (1991). The economic impact of agricultural
extension: A review. Economic Development and Cultural Change 39 (3): 607-
650.

Blundell, R and J. Powell (2004). Endogeneity in semiparametric binary response
models. Review of Economic Studies 71: 655-679.

Carney, D. (1998) Changing Public and Private Roles in Agricultural Service Provision.
ODI, London.

105

Chapman, R. and R. Tripp (2003). Changing incentives for agricultural extension. A
Review of privatized extension in practice. Agricultural Research & Extension
Network 132. ODI, London.

Christopolos, I. (1996) Poverty, Pluralism, and Extension Practice, Gatekeeper Series No.
64, International Institute for Environment and Development, London.

Christopolos, I. (2001) Extension, poverty and vulnerability in Nicaragua. Country study
for the Neuchatel initiative. Overseas Development Institute and Uppsala
University, London.

Deaton, A. (1997). The Analysis of Household Surveys. A Micro Econometric Approach
to Development Policy. The Johns Hopkins University Press, Baltimore and
London.

Eaton, C and A. Sheperd (2001). Contract farming: partnership for growth. Agricultural
Services bulletin 145 FAO, Rome.

Edwards, M. and D. Hulme (1996). Too close for comfort? The impact of ofﬁcial aid on
nongovernmental organizations. World Development 24(6):961-973.

F arrington, J ., A. Bebbington, K. Wellard and D. Lewis (1993) Reluctant Partners?-
Non-Governmental Organizations, the State and Sustainable Agricultural
Development. London: Routledge.

Farrington, J. (1994). Public sector agricultural extension: Is there life after structural
adjustment? Natural Resource Perspectives 2. ODI, London.

F arrington, J. and A. Bebbington (1994). From research to innovation: Getting the most
from interaction with NGOS in farming systems research and extension.
Gatekeeper Series 43. International Institute for Environment and Development,
London.

Femandez—Comejo, J ., S. Jans and M. Smith (1998). Issues in the economics of pesticide
use in agriculture: a review of the empirical evidence. Review of Agricultural
Economics 20(2): 462-488.

Feder, G., L. Lau, and R. Slade (1987). Does agricultural extension pay? The training and

visit system in northwest India. American Journal of Agricultural Economics
69:677-686.

F eder, G., R. Murgai and J. Quizon (2003). Sending farmers back to school: The impact
of farmer ﬁeld schools in Indonesia. Review of Agricultural Economics. 26(1):
45-62.

106

Fleischer, G., F. Jungluth, H. Waibel and J .C. Zadoks (1999). A ﬁeld practitioner’s guide
to economic evaluation of IPM. Pesticide Policy Project Publication Series No 9.
University of Hannover, Hanover.

Gallagher, K (1998). Farmer Field Schools (F FS) : A Group Extension Process Based on
an Adult Non-Formal Education Methods. F A0, Global IPM facility, Rome.

Garforth, C. and A. Lawrence (1997). Supporting sustainable agriculture through
extension in Asia. Natural Resource Perspectives 21. ODI, London.

Hasan, A. (1993). Scaling-up the OPP’s low-cost sanitation programme. Orangi pilot
project, Karachi.

Hashemi, S. (1992). State and NGO support networks in rural Bangladesh: concepts and
coalitions for control. Mimeo. Centre for Development Research, Copenhagen.

Hruska, A and M. Corriols (2002). The impact of training in integrated pest management
among Nicaraguan maize farmers: Increased net returns and reduced health risk.
International Journal for Occupational and Environmental Health 8: 191-200.

Hussain, S., D. Byerlee and P. Heisey (1994). Impacts of the training and visit extension
system on farmers’ knowledge and adoption of technology: Evidence from
Pakistan. Agricultural Economics 10: 39-47.

Khwaja, A. (2002). Can good projects succeed in bad communities? Collective action in
public good provision. Working paper. Harvard University, Cambridge, MA.

47pp.

Kenmore, P (2002). Integrated Pest Management. International Journal of Occupational
& Environmental Health 8(3):173-174.

Levard, L and Y. Marin (2000). La problematica técnica y la intervencion de los
organismos que trabajan en el fomento tecnologico en el tropico seco de
Nicaragua. Encuentro 32(53): 1 1-18.

Owens, T., J. Hoddinott and B. Kinsey (2003). The impact of agricultural extension on

farm production in resettlement areas of Zimbabwe. Economic Development and
Cultural Change 51(2): 337-357

Petrin, A. and K. Train (2005). Control function corrections for omitted attributes in a
differentiated product models. Working paper, University of Chicago, Chicago,

52 pp.

Piccioto, R. and J. Anderson (1997). Reconsidering agricultural extension. World Bank
Research Observer. 12(2):249-259.

107

PROMIPAC (Proyecto Manejo Integrado de Plagas en America Central) (2001)
Evaluacio’n de metas del proyecto PROMIPAC en su segundafase 1998-2001.
Esteli, Nicaragua.

PROMIPAC (Proyecto Manejo Integrado de Plagas en America Central) (2003) Informes
de las Escuelas de Campo de Agricultores de F rijol. Esteli, Nicaragua.

Qamar, K (2002) Global trends in agricultural extension: challenges facing Asia and the
Paciﬁc region. Food and Agriculture Organization of the United Nations URL:
http://www.fao.org[sd/2002/k_n0903a2ien.htm. Access date September 2, 2005.

Raina, R. (2003). Disciplines, institutions and organizations: impact assessments in
context. Agricultural Systems 78: 1 85-21 1.

Ravallion, M. (2005). Evaluating Anti-Poverty Programs. World Bank Policy Research
Working Paper N. 3625, June, Washington DC.

Reuchlin, P (2003). “Can NGO’s empower the poor and inﬂuence policy and planning?”
Working paper. London School of Economics, London.

Rivera, W. and D. Gustafson (1991). New roles and responsibilities for public sector
agricultural extension: the impact of multi-institutional activities. In Rivera, W.
and D. Gustafson (Eds) Worldwide Institutional Evolution and Forces for
Change. Elsevier, Amsterdam.

Rivera, W. (2001). The invisible frontier: the current limits of decentralization and
privatization in the developing countries: In E. Brewer (Ed). Agricultural
extension: An International Perspective. Erudition Press, North Chelmsford, MA.

Rivera, W. and W. Zijp (2002). Contracting for Agricultural Extension: International
Case Studies and Emerging Practices. Wallingford, UK: CAB International.

Rivera, W. and K. Qamar (2003). Agricultural Extension, Rural Development and the
Food Security Challenge. Sustainable Development Department. FAO, Rome.

Robinson, M. (1993). Governance, democracy and conditionality: NGOS and the new
policy agenda. In A. Clayton (Ed.). Governance, Democracy and Conditionality:
What Role for NGOS? INTRAC, Oxford.

Staver, C. and F. Guharay (2003). Building Integrated Pest Management practices in
Central America. In K. Maredia, D. Dkouo and D. Mota-Sa’mchez (eds) Integrated
Pest Management in the Global Arena, CABI international, Wallingford, UK.

Strauss, J., M. Barbosa, S. Teixeira, D. Thomas and R. Gomez (1991). Role of education

and extension in the adoption of technology: a study of upland rice and soybean
farmers in central-west Brazil. Agricultural Economics 5: 341-359.

108

Sulaiman, V. and A. Hall (2002). Beyond technology dissemination: Reinventing
agricultural extension. Outlook on Agriculture 31(4):225-233.

USDA (United States Department of Agriculture) (1998). Acute Toxicity of Pesticides.
Pesticide Data Program. URL:www.ecologic-ipmcom/PDP/Tablej1998.pdf.

Van der Ban, A (2000). Different ways of ﬁnancing agricultural extension. Agricultural
Research and Extension Network 106. ODI, London.

Vivian, J. (1994). NGOs and sustainable development in Zimbabwe: No magic bullets.
Development and Change 25: 181-209.

Wallace, 1. (1997). Agricultural education at the crossroads: present dilemmas and
possible options for the future in sub-Saharan Africa. International Journal for
Educational Development 17(1):27-39.

Wooldridge, J. (1997). On two stage least squares estimation of the average treatment
effect in random coefﬁcient models. Economic Letters 56: 129-133.

Wooldridge, J. (2002a). Econometric Analysis of Cross Section and Panel Data. MIT
Press. Cambridge, MA.

Wooldridge, J. (2002b). “Unobserved heterogeneity and estimation of average partial
effects.” Working paper. Department of Economics. Michigan State University.

Wooldridge, J. (2003). “Cluster sample methods in applied econometrics. ” Working
paper. Department of Economics, Michigan State University.

World Bank (1991). How the World Bank Works with Non-Governmental Organizations.
World Bank publications, Washington DC.

World Bank (2002). Summary report of the World Bank/USAID/Neuchatel international
workshop on Extension and Rural Development held at IFPRI. Washington DC.

109

CONCLUSIONS

Assessing the impact of agricultural extension in developing countries after the
major changes suffered by national extension services poses many challenges. Impact
studies should carefully examine whether targeted farmers perceive all potential beneﬁts
derived from technology adoption, measure all farm household outcomes associated with
the adoption process and carefully evaluate the impact of the new generation of non-
governmental extension providers. Great care should also be given to the methods for
estimating program impacts in order to avoid attributing impacts that do not correspond
to the extension program.

This dissertation has explored three major research questions related to the
delivery of integrated pest management (IPM) through the farmer ﬁeld schools (F PS), the
preferred means of many development agencies for disseminating IPM. These questions
are analyzed in three separate but related essays. The ﬁrst essay evaluates whether
awareness and/or experience of environmental and health (E&H) effects inﬂuence
farmers’ decisions on pest management. The second essay analyzes whether the
implementation of FFS has produced farm, environmental and health impacts among
participants in this extension program. Finally the third essay explores whether diverse
institutional characteristics of extension providers affect the impacts that extension
programs have on farmer clients.

Prior experience with the E&H effects of pesticides was found to affect household
decisions about pesticide use and the adoption of non-chemical pest control practices.
Perceived E&H hazards are only important enough to inﬂuence farmer practices when

farmers are using toxic or highly toxic pesticides. Below that level Nicaraguan bean

110

growers ignore potential E&H risks. Also when farmers face high health risks, farmers
tend to use protective strategies in order to avoid direct exposure. These ﬁndings partially
support the assumption by many IPM training programs that awareness of the E&H risks
of pesticide use will encourage a great IPM adoption. Few previous studies had tested
formally whether farmer perceptions of E&H risks inﬂuence their input decisions. IPM
training programs need to improve their educational component and stress the importance
of the E & H effects of pesticide use. Secondary effects that are not perceived directly by
farmers making decisions about pesticide use should also be incorporated in the IPM
training curricula.

Regarding the impact of F F8 for IPM training, this dissertation ﬁnds that failing
to correct for endogeneity effects associated with IPM training participation can
exaggerate the apparent impacts of FFS. Like a previous study-about FF S impacts, this
dissertation ﬁnds F FS performance to be inferior to other IPM training programs in
Nicaragua and to have created no diffusion effect among the neighbors of FF S graduates.
Also FFS did not produce the expected E&H beneﬁts associated with the delivery of
IPM. These results are partly explained by the poor agricultural resources (especially
poor soils and high pest infestation levels) in the hands of small farmers in developing
countries, the main target of F FS. However, the dissertation also ﬁnds that longer IPM
training programs achieve better outcomes. These ﬁndings highlight the importance of
using appropriate econometric methods to correct for potential endogenity of extension
program participation. There is also a need to expand traditional impact studies to include
the E & H effects associated with IPM adoption. Given the high cost of delivering IPM

through FF S and the limited impact achieved so far, it is important to adjust the FFS

111

approach. Cost savings can be achieved by prioritizing and trimming activities the current
FFS curricula. If these savings are invested in designing longer periods of IPM
dissemination to maintain and improve farmers’ feedback, FFS could improve their
outcomes.

Finally, in the search for explanations of the disappointing proﬁtability and
pesticide risk outcomes of FFS in Nicaragua, this dissertation explores the inﬂuence of
institutional characteristics among the NGOs in charge of implementing F FS. It ﬁnds that
NGO characteristics signiﬁcantly affect bean farmers’ choices of pest management
practices and to a lesser extent, their farm net revenues. In particular, the institutional
focus, expertise in IPM and the capacity of NGOs implementing multi-institutional
extension projects signiﬁcantly affect their clients’ input choices and adoption of
agricultural technologies. Previous impact studies have usually ignored the institutional
environment that could have hided the inﬂuence of individual institutions in the
aggregate treatment effect of extension programs. As the impact of extension programs
can be enhanced or limited by the individual characteristics of extension providers, there
is a need to improve selection of the institutions that deliver agricultural technologies,
focusing on strong technical capacity and an institutional emphasis that matches the main

focus of the extension program.

112

APPENDIX A.1

Household Survey

113

Encuesta sobre el uso de agroquimicos y métodos alternativos para el

control de plagas entre productores de frijol de Nicaragua.
Encuesta mimero: Fecha:

 

I. Informacién general

 

 

Nombre del jefe de hogar:

Ubicacion de la ﬁnca: Comunidad:

Municipio: Departamento:

C6digo de GPS: Punto No:

Lat: Long: Altura:

 

Nombre del encuestador:

II. Caracteristicas del hogar.
(Un hogar es deﬁnido por todos los miembros de familia que viven/trabajan
permanentemente en la casa/ﬁnca durante los siguientes ciclos agricolas.)

2.1 Composicion del hogar.

 

Total Menores de 14 Cuantos van a
aﬂos. la escuela

 

Varones

 

Muj eres

 

2.2 Informaeién del jefe de hogar.

2.2.1 Género (M=0, F=1)
2.2.2 Edad

2.2.3 Afios de educacion
2.2.4 Edad del conyuge
2.2.5 Aﬁos de educacién del conyuge

 

 

 

 

2.3 Aﬁos de educacion de los padres del jefe de hogar

2.3.1 Padre
2.3.2 Madre

2.4 Familiares que trabajan en el extranjero, marcar con una X.

2.4.1 Padres
2.4.2 Hermano/a
2.4.3 C6nyuge
2.4.4 Hijo/a
2.4.5 Otro:

 

 

 

 

114

2.4.6 Ninguno
Donde?

 

III. Activos del hogar y caracteristicas de la ﬁnca.

3.1 Cuantas manzanas tiene su ﬁnca.

 

3.1.1 3.1.2 3.1.3 3.1.4 3.1.5
Total Propia Alquilada Compartida Otra

 

 

3.1.6 Cuantas manzanas son para produccitin agricola

 

3.1.7 Cuantas manzanas tiene con sistema de riego

 

3.1.8 Tipo de riego.

 

3.2 Adquisicion de tierras (en manzanas).

 

3.2.1 Herencia 3.2.2 Comprada 3.2.3 Transferida 3.2.4 Invadida 3.2.50tro

 

 

3.3 Distribucién del area de cultivo y produccién durante el ﬁltima ciclo agricola (en
manzanas, kilos y c6rdobas)

 

 

 

Cultivo Ciclo , Produccién Cantidad Precio
agricola Area vendida unitario

Maiz

F rij ol

 

Maicillo

 

 

 

 

 

 

 

 

 

 

115

 

3.4 (A qué distancia se encuentra el lugar de venta de su produccion.
3 .4. 1Kilometros
3.4.2 Minutos

3.5 aCuanto le cuesta transportar el quintal de frijol al mercado?

 

3.6 aCuanto tiene de cada uno de los siguientes animales?

 

3.6.1 Vacunos 3.6.2 equinos 3.6.3 Cerdos 3.6.4 Aves 3.6.5 Otros
(aproximadam.)

 

 

3.7 Tiene electricidad? SI ( ) NO ( )
3.8 aCuales cosas de la siguiente lista tiene?

3.8.1 Carro____

3.8.2 Motocicleta

3.8.3 Televisor

3.8.4 Refrigerador

3.8.5 Moto bomba

3.8.6 Motor de riego_
3.8.7 Tractor

3.8.8 Otro

 

 

3.9 Trabajo por temporadas fuera de la finca.

 

Miembro Actividad Dias Salario / dia Distancia al lugar

de la ﬁnca Kms Minutos
Padre

Madre

Hijos

 

 

Nota: si no contesto e1 valor del salario preguntar :

 

3.10 Comparado al salario que paga en su finca, el salario fuera de ella es:

1. Mayor 2. Igual 3.Menor

116

 

£0 8 as:
.38 880

0.00

a
2&8
:00

«2600
am
8 on...

3
25 880 5.25

8580: \ 35
050 on 2.2: 0.. 002.55

no

0.80

a0
£09..—

.a
:00

N 00...»...—

880

m0
302.—

.0088
x 32020080

ave—mg
» 00.00.?

85.5..

N «200085
2. 593:?
_m20_83.0_
8 582.3.
003000

a 2.3280
«203.0:

8 508:?
$08.8".

3:58
2955

8.3.5.
050300.005

00808030
a 23.6

.050

33:2.— 103.53.

.—€_.—.-.u 0mg G—Umo Gama:— _0—w a—Ouhﬂn— ﬂ— :0 mﬂﬁﬂmﬂ «aha—O O—u GENE % 8Q¢Cm>mu0< —.V

3.0.... .8 8.3524:

117

Ahora me gustaria hablar un poco mas en detalle del uso de plaguicidas.
4.2 Use de plaguicidas. (Frijol)

 

Producto Numero de Frecuencia Capacidad de Dosis
aplicaeiones de aplic. la bomba. aplicada.

 

 

 

4.3 aCuales son las principales plagas que usted tiene normalmente en los frijoles? Y
gQué plaguicidas especificos usa normalmente para controlarlos

 

Ciclo agricola Postrera
Ciclo agricola Primera

 

4.3.1 Plaga Cod 4.3.2 Plaguicida Cod 4.3.3 Plaga Cod 4.3.4 Plaguicida Cod

 

 

4.4- Evaluacibn del nivel plagas y productos usados.

Respecto a un aﬁo promedio.

 

Plaga El Nivel Plaguicida El Nivel
plaga plaguicida USO de plaguicidas
fue: fue: Desde Plag. Usados cod
cuando anteriorrnente

 

N iveles: l-menor; 2-mismo; 3-Mayor; 9- No Sabe

118

 

4.5 Que métodos de control de plagas 4.5.1 Marcar con
que no requieran quimicos conoce? una X lo que usa.

 

 

 

4.6 Quién es responsable de aplicar los plaguicidas en su f'mca?
4.6.1 Jornal contratado
4.6.2 Hijos
4.6.3 Padre
4.6.4 Madre
4.6.5 Otros

 

V. Efectos en la salud por el uso de plaguicidas en frijol

Con relacion a los agroquimicos de los que estuvimos hablando recientemente, me
gustaria preguntarle sobre las plagas que estos controlan y sobre los efectos posibles que
podria producir su uso en la salud humana. Me gustaria saber su propia experiencia con
estos agroquimicos y otras formas de controlar plagas.

5.1 Conoce usted efectos que pueda producir el uso de plaguicidas en la salud?
Si () No ()

5.2 (gUsted o su familia ha suﬁ‘ido a1 gun problema en la salud a causa de los plaguicidas.
5.2.1 Si( ) 6 No()
5.2.2 Cuantas veces
5.2.3 Ha habido casos de intoxieacibn intencional Si () o No ()

En otras areas similares a esta region, ha habido reportes que muestran que el uso de

pesticidas puede causar efectos sobre la salud humana y sobre el medio ambiente. Nos

gustaria hablar sobre los efectos que usted y su familia ha experimentado por haber
estado expuestos directamente a los plaguicidas mientras cultivaban ﬁijol.

119

5.3 aCuales de los siguientes sintomas de intoxicaciones agudas han experimentado
los miembros de su familia? y gCuantas veces? (Marcar si experimenté en ultima
campana
Veces numero de miembros
5.3.1 Irritacion de los ojos
5.3.2 Escaldadura de la piel
5.3.3 Nausea y vomito
5.3.4 Dolor de cabeza
5.3.5 Mareo
5.3.6 Vision nublada
5.3.7 Molestias estomacales
5.3.8 Dolores musculares
5.3.9 Diﬁcultades para respirar
5.3.10 Diarrea
5.3.11 Otro

 

5.4 aQué hicieron para solucionar las intoxicaciones agudas en la salud que
experimento su familia? (Marcar la celda correspondiente).

 

 

SINTOMAS Doctor Clinica/ Curandero Auto- Nada Otro
local en la prescripcién
ciudad
Irritacion de los ojos
Escaldadura de la piel

Nausea y vomito
Dolor de cabeza
Marco

Vision nublada
Molestias estomacales
Dolores musculares
Diﬁcultades respiratoria.

 

Diarrea

 

5.5 Nos gustaria tener un estimado de los costos de salud y otros costos escondidos
que estan relacionados con la intoxicacion de pesticidas. LUsted recuerda cuanto
gasto 1a ultima vez que un miembro de su familia se intoxicé con plaguicidas? (En
moneda local).

 

Intoxicacion Servicios Medicina Dias de Otros Aﬁo
médicos trabajo de la
jerdidos intox

 

Gasto de recuperacion

 

 

120

5.6 Qué tipo de medicinas necesita para curar una intoxicacion?

 

 

 

5.7;D6nde almacenan los plaguicidas?

 

 

5.8 Usa de nuevo los envases de los plaguicidas? Si ( ) No ( )

5.9g,Para que?

 

 

5.10 gCuales de los siguientes implementos usa normalmente para protegerse contra la
intoxicacion de plaguicidas? (Antes y después).

 

Siempr Esporadi Nunca Costo Aﬁo de U86 ultimo

 

e camente (Cérd) compra ciclo gggcola
Mascara
Guantes
Ropa protectora

Primeros auxilios
Lavarse las manos.
Baﬁarse

Tomar leche

Tomar agua con tierra

 

 

VI. Efectos del uso de pesticidas en insectos benéficos en frijol.
Una de las formas mas importantes para controlar plagas es el uso de insectos benéﬁcos.
Me gustaria hablar sobre el uso potencial de ellos en su ﬁnca.

6.1 Sabia usted que algunos insectos son enemigos naturales de las plagas del ﬁ'ijol y
pueden ayudar a controlar estas plagas? Si () o No ( )

6.2 Si 6.1 no, Cree que algunos insectos puedan controlar naturalmente otros insectos?

121

6.35Qué insectos benéﬁcos conoce y que sabe de ellos?* (si responde NO pasar a la 7.1)

 

Insectos Cod Insectos que Cod Plantas Co Son susceptibles
controla reflﬂ'o d a plaguicidas?

 

99’3”?

 

6.3 gEn un aﬁo normal son suﬁcientes los insectos benéﬁcos que hay en su ﬁnca para
controlar las plagas que menciono? Si ( )o No ( ).

6.5 Si 6.4 es no. La presencia de estos benéﬁcos reduce el requerimiento de plaguicidas
en la parcela. Si( )0 No ( ).

6.6 Usted ha visto cuantos plagas puede controlar.
El insectol

El insecto 2

El insecto3

El insecto 4

6.7 Que hacen para conservar y atraer los insectos benéﬁcos.

 

 

 

6.8 En el ultimo ciclo agricola el nivel de insectos benéficos en el frijol fue.
6.8.1 Mayor 6.8.2 Menor 6.8.3 Igual

6.9 Dada la presencia de insectos benéﬁcos en el ultimo ciclo del ﬁ'ijol, el uso de
plaguicidas fue;

6.9.1 Mayor 6.9.2 Menor 6.9.31gual

122

VII. Conocimiento y adopcion de MIP y ECAs

7.1 Conoce el programa de Escuelas de Campo para Agricultores. Si ( ) no ( ).
7.1.1 5Participa en una de ellas?

7.1.2 Conoce alguien que participo. Si ( ) no ( ).

7.1.3 Escucho hablar de ellas

7.1.4 En qué aﬁo tuvo su primer contacto?

7.1.5 No conoce—

7.2 Si conoce.
7.2.1 En qué cultivo de la ECA
7.2.2 En qué aﬁo tuvo su primer contacto.

7.3 Conoce el MIC Si( ) y NO ()

 

7.4 Conoce el MIP Si ( ) y NO ()

 

7.5 Qué actividades de MIP / MIC esta im alementando en su ﬁnca?

 

 

Area Are
MIP MIC a
7.5.1 Analisis de agroeco- 7.5.11 Incorporacion de
ecosistema. rastrojos
7.5.2 Recuento de 7.5.12 Arado en seco
plagas/enferm
7.5.3 Trampas 7.5.13 Abono verde
7.5.4 Conservacibn de insectos 7.5.14 Acequias
benéficos.
7.5.5 Plaguicidas botanicos. 7.5.15 Agroforesteria
7.5.6 Plaguicidas biolbgicos 7.5.16 Barreras vivas
7.5.7 Proteccibn de semillero. 7.5.17 Barreras
muertas.
7.5.8 Agua caliente 7.5.18 Terrazas.
7.5.9 Ca1+ ceniza 7.5.19 Curvas a Nivel
7.5.10 Control de malezas 7.5.20 Prueba de
hospederas germinacibn
7.5.21 Abono organico/
biofertilizante

 

7.6 Qué cambios Experimentb en su finca después de implementar MIP?

 

 

 

123

7.7 éComo cambio el uso de plaguicidas en frijol la adopcion de MIP?

 

Incremento (litros) Decrecio (litros)

 

Plaguicidal
Plaguicida2
Plaguicida3

 

Plaguicida4

 

7.8 aCbmo cambiaron sus rendimientos de frijol con la adopcibn de MIP?
(Quintales).

7.8.1 Se incremento en
7.8.2 Decrecio en

7.9 Como cambiaron los requerimientos de mano de obra con la adopcibn de MIP?

7.9.1 Se incrementé en jornales
7.9.2 Decrecieron en jornales

7.10 aComo cambio el uso de otros insumos en su produccién de frijol con la
adopcién de MIP?

 

Insumo Nunca uso Incremento (%) Decrecio (%)

 

1.

 

2.
3.
4

 

7.11 Mejoro su conocimiento sobre la dinamica de insectos?

7.11.1 Tengo el mismo nivel
7.11.2 Mejoré mi conocimiento
7.11.3 conoce menos

7.11.4 No sé

7.12 Trabaja con alguna de las siguientes instituciones?

7.12.1 PROMIPAC 7.12.7 UNICAM_
7.12.2 INTA 7.12.8 CARE

7.12.3 CARITAS 7.12.9 UNAG

7.12.4 FIDER 7.12.10 ODESAR_
7.12.5 ESETECA 7.12.11ADAAC
7.12.6 CECOTROPIC 7.12.12 Otros

124

7.13 Que tipo de apoyo recibe de éstas instituciones?

Medias Fines
7.13.1 Asistencia técnica __ 7.13.6 Conservacion de suelos
7.13.2 Capacitacion _ 7.13.7Reforestaci6n
7.13.3 Crédito _ 7.13.8 Salud
7.13.4 Alimentacion _ 7.13.9 Enfoque de genero—

7.13.5 Pago directo / incentivos_ 7.13.10 Organizacion
7.13.11 Mejoramiento de ganado_

7.14 Tiene usted acceso a crédito a partir de:

7.14.1 Familia
7.14.2 Amigos
7.14.3 Proveedor
7.14.4 Proyecto
7.14.5 Banco

7.14.6 No tiene crédito_

 

 

 

 

7.15 OBSERVACIONES.

 

 

 

 

125

APPENDIX A.2

Map of Nicaragua

126

90°U'U‘VV 85°D'U'W

Map of Nicaragua

2U‘D‘D'N 20'D'U'N

Guatemala

15'0'0'N 15'D'U'N

El Satvador

E] Limits
Nicaragua
10'0'0'N

lﬂ'ﬂ'ﬂ'N

Panama

 

127

APPENDIX A. 3

Weighting scheme to correct sample stratiﬁcation

128

Weighting scheme

We deﬁne Q, = Pr(Q € Wj) as the population frequency for stratum j (probability that
each observation belongs to stratum j) and Hj = Nj/N as the current proportion of each
stratum within the sample. Then the weight for the stratum j can be deﬁned as Qi/Hj. This

weighting scheme is incorporated in the survey regression methods

129

 

F ‘4‘ 'mznﬁ

APPENDIX A.4

Potential measurement error in Essay 1’s environmental and health variables

130

Potential measurement errors in environmental and health variables

It is possible that respondents misreported the true acute health symptoms suffered by
household members or over reported the true level of observed beneﬁcial insect
population, in order to obtain additional beneﬁts after participating in IPM training. We

can represent these variables as follows:

H0=HR+uH

Bo=BR+HH

Where the observed health and beneﬁcial variables equals the real health and beneﬁcial
variables plus the measurement error (11) in each case. The circumstances under which
measurement error can produce biased estimates depend on the assumption about
correlation between the measurement error and the observed and real explanatory
variables. If we assume that the measurement error is uncorrelated with the real variable
(Cov[Xo,ui]=O), we are assuming that it is correlated with the observed variable, because
HFXo-XR (Wooldridge 2002). As we will estimate a model including the observed
variables, this assumption would create an endogeneity bias in our estimation and
instrumental variable estimation would be necessary to correct the problem. If we assume
that the measurement error is uncorrelated with the observed variables (Cov[XR,ui]=O),
we are assuming that it is correlated with the real variable. In our case the second
assumption seems to be more plausible. It is very likely that households that did not

suffer from acute health effects or that observed no beneﬁcial insects will report zero

131

incidence, and therefore the measurement error for them is zero. But households that
have suffered acute symptoms are likely to overreport these symptoms in order to give
the researcher the impression that they need extra help from externally funded projects,
and households that have observed beneﬁcial insects are also likely to overreport their
levels in order to give the impression that they beneﬁt from previous training and are
available for collaboration. This implies that the measurement error is correlated with the
true variables. In this case we can still produce consistent estimates but error variance

will be increased

132

APPENDIX A.5

Production and Socioeconomic Variables used in all regression models

133

Production and socioeconomic variables used in econometric analysis

 

 

Variables Variable codes Mean Std Dev.
Production Variables

Price of beans (USS per kilo) BEANPR 0.3 0.1
Price of maize (USS per kilo) MAIZEPR 0.1 0.0
Price of bean seed (USS per kilo) SEEDPR 0.5 0.3
Price of metamidophos (USS per liter) IN SECPR 4.6 0.8
Price of gramoxone (USS per liter) HERBPR 5.7 0.7
Price of mancozeb (USS per kilo) FUNGPR 4.2 0.6
Price of fertilizer (USS per sack of 1001b) F ERTPR 12.2 2.5
Price of oxes (USS per day) OXESPR 12.7 4.3
Wage for spraying chemicals (USS per man-day) WAGE 1.8 0.4
Transport cost (USS per sack of 100 lbs) TRANSP 0.6 0.3
Distance to municipal capital (Kms) MUNDIST l 1.3 7 .9
Farm altitude (meters above sea level) ALTITUDE . 762.9 232.2
Postrera farming season (%) POSTRERA 84.0

Primera farming season (%) PRIMERA 10.0

Apante farming season (%) APANTE 6.0

Households that observed high level of insect pest (%) HIGHPEST 8.0

Households that observed high level of diseases (%) HIGHDIS 8.0

Households that observed high level of slugs (%) HIGHLUGS 9.0

Households with IPM training (%) IPMTRAIN 62.8

Household characteristics

Households with female head (%) FEMHEAD 15.0

Age of household head (years) AGEHEAD 45.0 13.8
Education of household head (years) YEAREDU 3.2 3.2
Proportion of female (%) FEMPERC 49.0

Number of children NCHILD 2.2 1.7
Total area of land (hectares) TOLAND 8.4 12.5
Households with electricity at home (%) ELECTRIC 44.5

Number of cattle heads CATTLE 3.6 6.7
Household receive remittances ﬁ'om relatives (%) REMITTA 21.0

 

134

APPENDIX A.6

Acute health symptom frequencies used in Essay 1

135

 

'- .'. .‘le‘. "isn't!

Frequency of acute health symptoms (in percentage) for 436 Nicaraguan bean

 

 

_growers, 2003-04
Never Rarely Sometimes Frequently
Headache 51.8 17.2 12.2 18.8
Eye irritation 56.9 14.2 12.6 16.3
Stomachache 76.6 9.4 6.0 8.0
Skin rash 71.0 12.9 8.3 7.8
Blurred vision 72.4 11.7 7.4 8.5
Muscle pain 73.8 9.7 8.7 7.8
Respiratory difficulties 88.8 3.9 3.2 4.1
Diarrhea 96.8 0.7 1.8 0.7

 

136

APPENDIX A.7

Complete regression results in Essay 1

137

I

Complete regression results for the pesticide demand+ survey least squares
rggression for 436 Nicaraggan bean growers, 2003-04

 

 

 

Insecticide Herbicide Fungicide _ Molluscicide
Bean Price -0.000 -0.001 0.001* -0.000
(0.59) (1.05) (1.93) (0.45)
Seed Price -0.012 -0.007 -0.012 -0.012
(0.89) (0.48) (0.96) (0.43)
Insecticide Price -0.013*** 0.002 0.005
(3.00) (0.73) (1.15)
Herbicide Price -0.043***
(5.00)
Transport cost -0.001 -0.010 0.006 0.003
(0.05) (0.90) (0.90) (0.36)
Wage -0.004 0.003 0.000 0.001
(0.48) (0.37) (0.07) (0.10)
Altitude 0.000 0.001 -0.002 0.000
(0.28) (1.23) (1.09) (0.49)
High pest level -0.059
(0.40)
high level of diseases 0.442***
(3.47)
high level of slugs l.027***
(3.06)
Municipal distance -0.000 -0.256** -O. 120 -0.088
(0.00) (2.09) (1.40) (1.18)
Adopted AAEH -0.093 -0.007 -0.007
(0.63) (0.07) (0.08)
AAE & more beneficials -0.171 0.337 -0.050
(1.16) (1.32) (0.25)
Adopted botanicals insect. -0.269 0.154" -0.135
(1.59) (2.08) (0.89)
Adopted yellow traps 0.008
(0.05)
1PM trained 0.339*** 0.151* 0.060 0.070
(2.96) (1.66) (0.78) (0.65)
IPM & beneﬁcial awareness -0.336* -0.059 0.007
Number of symptoms 0.012 0.070 -0.003 -0.010
(0.37) (1.37) (0.14) (0.39)
Freq. eye irritation -0.009 -0.027 0.002 0.002
(0.15) (0.47) (0.04) (0.04)
Freq. diarrhea -0.398* -0.048 0.01 3 -0. 1 1 8
(1.74) (0.47) (0.13) (1.01)
Freq. respiratory difﬁculty 0.084 0.012 -0.020 -0.055
(0.86) (0.14) (0.3 5) (0.62)
Freq skin rash -0.106* -0.028 -0.017 0.013
(1.65) (0.52) (0.42) (0.28)
Freq. Stomach ache 0.039 0.035 -0.067 -0.048
(0.54) (0.33) (1.58) (0.62)
Freq head ache 0.177“ 0.002 -0.050 -0.029
(2.26) (0.03) (1.41) (0.60)
Freq muscle pain 0.052 -0.012 0.088 -0.040
(0.87) (0.21) (1.47) (0.74)
Freq blurred vision -0.017 -0.137* -0.029 0.060

 

 

 

 

 

 

 

(0.16) (1.98) (0.56) (0.61)
Lost workdays -0.233* -0.033 -0.000 -0.040
(1.65) (0.30) (0.00) (0.43)
Visited city doctor -0.022 0.078 0.051 -0.031
(0.13) (0.54) (0.39) (0.26)
Hospitalized -0.128 0.241 0.068 -0.053
(0.52) (1.53) (0.40) (0.39)
Purchased protective devices 0212*" -0.046 0.028 -0.003
(3.57) (0.85) (0.78) (0.08)
Homemade protective devices -0.l78* -0. 146 0.039 -0.034
(1.73) (1.49) (0.45) (0.41)
Hired applicator 0.189* 0.222* 0.082 0.104
(1.89) (1.72) (1 .06) (1.08)
Apante season -0.034 -0.292 0.172 0.101
(0.11) (1.00) (1.11) (0.30)
Postrera season -0.083 0.049 0.007 -0.042
(0.50) (0.26) (0.07) (0.14)
Female head 0.136 ~0.239 0.208* -0.138
(1.02) (1.63) (1.85) (1.55)
Years of education 0.018 0.024* 0.01 1 0.019
(1 . 16) (1.68) (0.90) (0.96)
Remittances 0.010 0.026 0.110 0.326"
(0.09) (0.22) (1.24) (2.01)
Total land -0.001*** 0.001*** 0.000 -0.000
(3.61) (3.08) (1.41) (0.82)
Number children -0.062** -0.018 0.050*** -0.01 1
(2.26) (0.63) (2.96) (0.42)
Female percentage -0.3 78 0.550* -0.428*** -0.180
(1.37) (1.94) (2.68) (0.81)
Has electricity 0.059 0.135 0.039 -0.217
(0.50) (1.04) (0.34) (1.35)
Cattle heads 0.005 -0.012 -0.001 0.003
(0.66) (1.10) (0.10) (0.39)
Constant 1.714 6.450*** 2.553 0.451
(1.22) (3.41) (1.12) (0.41)
Resistant variety 0.147
(1.38)
Observations 436 436 436 436
R-squared 0.46 0.52 0.37 0.39

 

‘ Signiﬁcant at 10%; "”' signiﬁcant at 5%; ‘”” signiﬁcant at 1%
Absolute value of t statistics in parentheses

I 73 community dummy variables are not reported in this table due to space limitation

1* Agro-ecosystem analysis

 

139

Complete regression results for the IPM adoption+ survey multinomial logit
regression for 436 Nicaraggan bean growers, 2003-04

 

 

Agro-eeol Botanical Insect Two IPM Three
Analysis Insecticides traps practices practices
Seed Price -0.083 -0.194 4.088 0.141* -0.048
(0.47) (1.34) (0.91) (1.67) (0.32)
Bean Price -0.007 0.001 0.199 -0.006 -0.003
(0.98) (0.30) (1 . 17) (1.59) (0.41)
Maize Price -0.011 0.026*** 0.731 0.028“ 0.011
(0.64) (3 .42) (1 .22) (2.5 1) (0.82)
Insecticide Price 0.045* -0.005 0.758 -0.049 0.013
(1.85) (0.23) (1.12) (1.51) (0.47)
Herbicide Price -0.058 -0.022 -0.500 0.017 0.065"
(1.41) (0.76) (1.03) (0.65) (2.55)
Transport cost 0.055 0.055 3.029 0.078“ -0.327***
(0.87) (1.37) (1.07) (2.01) (3.42)
Wage 0.161“ -0.019 1.084 0.114* -0.015
(2.24) (0.31) (1 .08) (1 .85) (0.17)
Altitude 0.002 0.000 -0.019 0.004“ 0.001
(1.09) (0.19) (1.23) (2.29) (0.58)
Municipal distance -0.412** 0.038 -4.710 0.029 0.078
(2.22) (1.11) (1.15) (0.73) (1.38)
1PM trained 4.793*** 5.661*** 24.656 8.624*** 89.649***
(4.10) (7.55) (1.61), (3.89) (14.03)
Number of symptoms 0569*“ 0.768*** 5.127 0.179 0.053
(2.77) (3.56) (1.44) (0.75) (0.20)
Freq. eye irritation 0.268 -0.757** -6.854 -0.058 0.467
(0.54) (2.19) (1.10) (0.22) (1.54)
Freq dizziness -0.243 -1.880*** 19.919 -0.640* -0.566
(0.53) (3.67) (1 . 16) (1.78) (1.28)
Freq. respirat. difficulty 0.021 0.891 ~18.574 1.995*** 1.347
(0.03) (1.57) (1.29) (3.20) (1.52)
Freq skin rash 0.894* 0.032 11.439 0.162 l.027***
(1.75) (0.09) (1.10) (0.36) (3.08)
Freq. Stomach ache 0.449 -0.314 12.612 0.007 0.182
(1.18) (0.81) (1 .07) (0.02) (0.39)
Freq head ache -0.290 0.630" 4.179 0.420 0.817“
(0.67) (2.41) (1.06) (1.45) (2.58)
Freq muscle pain 0.304 -0.553 12.274 -0.681 -0.302
(0.44) (1.41) (1.12) (1.55) (0.75)
Freq blurred vision -0.900 -1.100** -10.756 -0.845 -1.651**
(1.50) (2.22) (1.14) (1.59) (2.59)
Lost workdays -4.032* * -0.405 -33 .5 18 -0.810 -0.290
(1.99) (0.57) (1.28) (0.88) (0.36)
Visited city doctor 3.377 -1.105 -8.263 -2.079 -5.862***
(1.63) (0.84) (1.05) (1.42) (3.01)
Hospitalized 47.373*** 0.767 8.726 -0.861 42.351***
(15.34) (0.66) (0.86) (0.64) (19.94)
Purchased protective 0.785* 0.197 18.881 0.295 0.244
devices
(1.95) (0.74) (1.23) (1.20) (0.5 8)
Homemade protective 1.699 0.551 14.538 0.554 1.553“
devices
(1.53) (1.16) (1.31) (1.00) (2.29)
Hired applicator -0.421 1.237“ -52.568 -0.827 -0.882

 

140

 

 

(0.33) (2.00) (1.27) (1.16) (0.91)
Apante season 23.663 -2.084 345.235 -3.131 1.880
(0.65) (0.82) (1.03) (1.38) (0.80)
Postrera season 0.503 -1.000 119.253 0.868 0.840
(0.57) (1.03) (0.53) (0.85) (0.65)
Female head -0.203 -1.431* 44.229“ -0.753 0.101
(0.14) (1.92) (4.99) (0.87) (0.10)
Years education -0.062 0.104 -3.058 0.200“ 0.420***
(0.58) (1.30) (1.08) (2.31) (4.88)
Remittances -2.217* -0.171 19.782 0.605 0.976
(1.84) (0.28) (1.11) (1 .06) (1.27)
Total land -0.104** -0.010 -1.671 0.003 0.048"
(2.16) (0.67) (1.15) (1 .06) (2.11)
Number of children -0.320 -0.309* 9.016 0236 -0.303
(1.26) (1.88) (1 .25) (1.52) (1.09)
Female percentage 0.404 1.005 -24.828 3.173* 2.386
(0.33) (0.79) (0.89) (1.74) (1.40)
Has electricity 0.344 -1 .091 57.058 -0.850 -0.497
(0.48) (1.45) (1.23) (1 . 17) (0.60)
Cattle heads 0.023 0.098 6.028 0.101* 0.119
(0.28) (1.60) (1 .22) (1.66) (1.24)
San Isidro municipal.+ 7.081*** -0.688 -70.505** -0.703 -1.424
(2.86) (0.53) (2.34) (0.35) (0.61)
Wiwili municipal. -18.245*** 1.682 -202.242 6.086“ 0.563
(10.20) (0.60) (0.99) (2.43) (0.24)
Pueblo nuevo municipal. 1.765 3.083*** -43.822 2.907 ** 2.353
(0.92) (2.67) (1.19) (2.17) (1.60)
San Nicolas municipal. 1.766 -0.041 1.015 -0.374 -2.567
(0.89) (0.04) (0.20) (0.39) (1.44)
San Marcos municipal. 0.945 1.037 -122.748* 0.203 1.121
(0.43) (0.76) (1.78) (0.12) (0.66)
Condega municipal. 4.067" 1.094 -7.892** 1.838 2.561
(2.56) (1 .07) (2.02) (1 .30) (1.65)
Placaguina municipal. 1.436 1.003 -105.794* -1.253 0.933
(0.64) (0.56) (1.84) (0.67) (0.47)
Niquinohomo municipal. 2.737 -0.104 -72.562 2.560 0.967
(1 .28) (0.06) (1.20) (1.54) (0.52)
La Trinidad municipal. 1.522 -0.914 -10.110** 44.741 *** -10.210***
(0.71) (0.76) (2.35) (20.04) (2.76)
Constant -8.499* -6.637* -392.582 -18.984*** -102. 160
(1.69) (1.78) (1 .01) (2.97) (0.97)
Observations 436 436 436 436 436

 

"' Significant at 10%; ” signiﬁcant at 5%; "" signiﬁcant at 1%

Absolute value of t statistics in parentheses

I The municipality of Esteli is used as the control group

 

141

 

Complete regression results for the labor demand+ survey least squares regression
for 436 Nicaraguan bean gowers, 2003-04

 

 

 

Famm labor Hired labor Total labor
Log Seed Price -0.769 0.787 -0.467
(1.34) (1.23) (0.83)
Log Bean Price 1.165 -1.141 0.496
( 1.58) (1 .08) (0.79)
Log Insecticide Price 2.424* -0.260 0.976
(1.90) (0.12) (0.85)
Log Herbicide Price -3.554* -1.706 -4.697**
(1.74) (0.86) (2.42)
Log Wage 1.084 -0.261 0.944
(0.97) (0.18) (0.95)
Altitude -0.001 0.004 -0.001
(0.54) (1.10) (0.27)
Municipal distance 0.1 19 -0.016 0.228
(0.24) (0.02) (0.57)
Adopted ABA 0.432 -0.060 -0.1 18
(0.68) (0.08) (0.23)
AEA & more beneﬁcials -1.872** -0.269 -1.636**
(2.36) (0.20) (2.12)
Adopted botanical insecticides -0.959 0.615 -0.751
(1.63) (0.81) (1.49)
Adopted yellow traps 0.594 -2.422*** 0.217
(0.91) ’ (2.77) (0.39)
IPM trained 0.910" 0.031 0.638
(2.17) (0.04) (1.51)
IPM trained & beneﬁcial awareness -l.856*** 0.77 8 -0.894
(2.87) (1.12) (1.63)
Number of symptoms 0078 0.253 0.015
(0.59) (1.32) (0.15)
Freq. eye irritation -0.277 0.640“ -0.180
(1 .20) (2.02) (0.96)
Freq. diarrhea -0.43 1 1.032 -0.886*
(0.58) (1.33) (1.67)
Freq. respiratory difﬁculty 0.095 0.548 -0.163
(0.28) (1 .22) (0.52)
Freq skin rash -0.169 -0.689* -0.371
(0.64) (1.86) (1.52)
Freq. Stomach ache 0.158 -0.496 0.113
(0.54) (1.11) (0.45)
Freq head ache 0.163 -0.537 0.046
(0.77) (1.60) (0.26)
Freq muscle pain 0.346 -0.163 0.245
(1.64) (0.41) (1.33)
Freq blurred vision -0.620*** -0.248 -0.259
(2.67) (0.48) (1.46)
Lost workdays -0.542 -0.766 -0.543
(0.80) (1 .06) (1 .00)
Visited city doctor -0.981 -0.144 -0.813
(1.02) (0.16) (0.97)
Hospitalized 0.515 -1 .149 0.224
(0.65) (0.84) (0.32)
Purchased protective devices 0.426* -0.287 0.099
(1.70) (1 .02) (0.43)

 

142

 

Homemade protective devices -O.712 -0.316 -0.765*
(1.56) (0.61) (1.81)
Apante season -3.066*** 0.734 -0.956
(2.73) (0.29) (0.62)
Postrera season 0.814 0.725 0.856
(1.40) (0.77) (1.53)
Female head -1.398** 1.047 -0.449
(2.54) (1.21) (1.09)
Years of education 0.090 0.279*** 0.154***
(1.21) (3.79) (2.89)
Remittances 0.052 0.604 0.408
(0.10) (0.87) (0.91)
Log Total land -0.022 0.736“ 0.119
(0.09) (2.54) (0.59)
Number children -0.025 -0.073 -0.048
(0.19) (0.45) (0.43)
Female percentage -0.347 -0.637 0.349
(0.35) (0.34) (0.36)
Has electricity -0.863 0.413 -0.100
(1.19) (0.54) (0.15)
Cattle heads -0.084** 0.063 -0.007
(2.06) (1.11) (0.34)
Constant 2.984 -5.682 12.156
(0.22) (0.35) (0.98)
Observations 436 436 436
R-squared 0.43 0.37 0.46

 

"' Signiﬁcant at 10%; " signiﬁcant at 5%; *" signiﬁcant at 1%
Absolute value of t statistics in parentheses
I 73 community dummy variables are not reported in this table due to space limitation.

 

143

APPENDIX A.8

Complete regression results in Essay 2

 

9"»!-

144

Hausman test for the linear approximation of health function models

 

 

F-statistic P-value
Respiratory difﬁculties 0.77 0.5471
Eye imitation 0.90 0.4687
Stomachache 1 .73 0.1479
Blurred vision 0.46 0.7628

 

145

Predicted probability of FFS and other IPM training participation used as
instruments in the ZSLS estimations, survey probit regression for 436 Nicaraguan
bean growers, 2003-04+

 

 

FFS Other IPM program
Original explanatory variables
Bean price 0.000 -0.004"”"
(0.05) (2.07)
Insecticide price 0.001 0.000
(0.12) (0.03)
Herbicide price -0.001 0.012
(0.07) (1.13)
Fertilizer price 0.000 0009""
(0.05) (3.14)
Seed price 0.043 0.031
(1.39) (0.97)
Transport cost -0.025 0.022
(0.77) (0.89)
Wage -0.035"”" -0.014
(2.22) (0.91)
Altitude 0.002" -0.002"
(2.04) (2.52)
High pest level 0551" 0.614‘
(1.75) (1.76)
High disease level 0.136 -0.883”
(0.33) (2.47)
High slug level -0.246
(1 .00)
Municipal distance -0.055 0.039
(1.49) (1 .34)
Visit city doctor -0.89l "”" 0.404
(2.08) (1.04)
Protective devices 0.240“ ~0.068
(2.05) (0.60)
Hired applicator -0.315 0.055
(1.30) (0.19)
Apante season -1.391"' 1.631’
(1.82) (1.93)
Postrera season 0.086 1213‘"
(0.28) (3.46)
Female head 1.106‘" -1.209"‘"
(4.15) (3.44)
Years of education -0.018 -0.010
(0.67) (0.30)
Remittances -0.048 0.256
(0.18) (1 .30)
Total land 0.008 -0.004
(0.88) (0.30)
Number of children 0.062 0.047
(1 . 17) (0.82)
Female percentage -2.020"‘" 0.747
(3.25) (1 .41)
Has electricity 0.710" -0.339
(2.50) (1.25)
Heads of cattle -0.032 0.023

 

146

 

(1.33)

Redundant variables
Number of soil practices 0.007
(0.07)
Adopted terraces 0.121
(0.49)
Adopted countour -0.360
(1.38)
Ad0pted fences 0.529‘
(1 .85)
Adopted other soil practices 0.231
(1.01)
Received food from NGO 0.505"
(2.51)
Received training from NGO 0.173
(0.75)
Received credit ﬁom NGO 0.447 "
(2.23)
Constant -2.963"
(1.82)
Observations 436

(1.32)

0.016
(0.14)
0.601”
(2.24)
-0.291
(1.02)
0.178
(0.61)
-0.228
(0.80)
.0343
(1.20)
0.258
(1.09)

-0.604”"'

(2.73)

-63 .029‘ "'

(2.05)

436

 

“ Signiﬁcant at 10%; "“" signiﬁcant at 5%; ‘” signiﬁcant at 1%
Absolute value of t statistics in parentheses
+ Community dummy variables not reported due to space limitations

 

147

 

 

148

8...? .8? .8? 8...? 8...? 8...? 8.8.38 82.:
.8.... .2... 8.... .8... .8.... .3...
8...? 8.... 8.... . .8? 8.... 8...? .838 8.822.. ”Ransom
.2... .8.... .8.... .8. .. .8.... .8. ..
8.... 8.... .8... .8... 8.... m8... .838 2.8.2.. 88.2....
.8.... .8.... .8.... .3... .8.... .8....
8...... .8? .8? .8... 2...? 8.... 88:88.
.8.... .8.... 5.... .8. .. .8.... a...
8.... 8...... 8.... .8...? N8... 8.... 8.8....em .3... 28>
E... .8.... .8. .. .8. .. .8. .. .8....
.8? .8? .88.? .8...? 8.? 8.... 2... .28 a m"...
.8. .. .8.... .8.... .8... .8.... .8....
8.... 8.... .8... .8... 8...? 8.... 8.58 2... .050
.8.... .8.... .8.... a... .. .8.... .88
.8... .8... 8...? 8.... 8.? .88.... m...
.2. .. .8.... .8.... .88. an... .28
L8... .8... .8? ...8...? 8...? .8...? 8:8... .8882
.8.... .8....
N8... 8...? .8... v.8 a»...
.8.... .40... .8.... .8.... .8.... .2...
..8... 8.... 8.... 8.... 8...... 8.....- 83:...
.8.... .8.... .8. .. .8.... .2... .8....
8.... 8.... 8.... 8.... 8.... 8...? 083
.2. .. .3... .8.... .8.... .3... an...
8.... 8.... 8.... 8.... 8.... ..8... a8 8&8:
.8.... .8.... .88 .8. .. .8.... .8....
8.... 8.... .18.? . .8? .8? 8.... 8.... 88.
.8. .. a... ..
.8...? .8...? 8.... 8.0.8:".
.2... .8....
2.02561 8.366661 oota 0303.5:
8.... 0.2.
8...? : .8? 8.... 8.888.
.8. .. .8. .. .8.... .8.... .8. .. .8. ..
.8.... 8.... 8...? 8...? 8...? 8...? 8.... :80
323?: >_ 020%.: 8205.5: >— 822801 820.88.: >— 032.com:—

 

 

velmgu 8.330% :00: unawaaaom Z emu he 008.».an.— 800050 .000— hotsm +2552. 0233.0.— ..8 8.0.3.. nemmmoawma 30—9000

 

3.3.8: 32% o. 2... 035 25 a. v2.2.2 .2. o... 832...; >883. b.5888 Q. +

8355...... 5 8.838.. . .«o 022., 328.7.

2... 2 .582... :. a... a 2.2.2... : .8. ... .582... .

 

am...
$9.

.8...
2...?
.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....

9.6
emv

.8...
2...?
.2...
8....
.8...
8....
.8....
8...?
.8....
8...?
.8....
8....
.8....
8....
.8....
8....
.8....
8....
.8...
8....
.8...
.8...
.8...
8....
.8...
8....
.8...

2...
09.

.8...
88...

.8....
8...?
.8...
8....
.8...
.8....
.8....
8...?
.88
3.256
.8...
8...?
.8....
8....
.8...
.8...?
.8....
8....
.8....
8...?
.8....

NV...
09.

.88

2......”«66

.8...
8...?
.8....
8....
.8...
8....
.8....
8...?
.88

2.2.5.5...

.2...
8....
.8....
8....
.8...
8.....-
.8....
8....
.8....
8...?
.8....

N2.
one

.8...
8....

.8...
8...?
.8....
8....
.8....
88...
.8....
8...?
.88

3.256

.8....
8...?
.8...

.8...
.2...

.8...
.8...
8...?
.8...
8...?
.2...

cm...
on...

.88

Eula—d

.8....
8...?
.8...
8....
.8...
8...?
.8...
8...?
.8....

2332.6

.8....
8...?
.8...
8....
.8...
..8...
.8...
8...?
.88

388....

.2...

@0832...
.8.....ng

2.82.00

_o>o_ 828:. am...
but? .5328...
2.80 we 2.8...
$2.820 8:
2280.. no 0323......
5.220.... 85:52
2.2 :30...
80.83.83—
5.82...» we 28>
can: 2an".
:88. 828....

.888 8.8....

149

 

 

 

~88 ~88 ~88. ~88. _88 _88 888. .88. 883. 82.8..
5.8 8.8 88.8 :88 a8. 8 an. 8 5. 8 38.8
888- 288. 3:8. 8 _8. .888. 8~8. 83. «~88. seam 38%
2 .8 $8 5.8 8.8 68.8 88.8 5.8 E .8
2.888. :88. 2:328 5228 $88. _88. 888 288 .9888 82.:
88.8 3.8 38.8 c .8 98.8 88.8 808 5.8
88.8 388 8 _ 8. 888 888 888. 8:8. 888 8888» 58 38>
G~8 5.. 8 G~8 28.8 8.8 28.8 5.8 :8. 8
28.8 .88. ~88 888. 88 ~88 «88 ~28. 882.85
88.8 5.8 88.8 :88 A88 2 m. 8 c _8 5.8
83 8 _ 88 38 ~88 :88 888. m8... 2 _8. 2.: .28 85 wt
8.8 8.8 38.8 5.8 A28 38.8 5.8 88.8
aﬁavmé 2.3—:6 2:536 Sikamd oﬁ—wmé Sikh—.6 SiwaA ioumacd 2a: .550
5.8 an. 8 5.8 88.8 ¢~8 $8.8 $8.8 88.8
8 .8 8.8 :88 :-m8 828 .2888 888 58:; m“;
$8.8 :88 A8. 8 88.8 88.8 88.8 8 08 8.8.8
288 ~88 L _ _8. 888. 88. _~88. .28. 88. 8888 .2828:
an. 8 8.8 c _.8 A88 :88 5.8 A8. 8 38.8
. 2:8. .888. :88. $88 ~88. _88. 8.8. 2 _8. 3888 8.:
$88 98.8 8.8 88.8 5.8 88.8 5.8 38.8
888. 88. ~88. _88 888. 888 888. 888 v.8 :8:
£88 $8.8 $8.8 $~8 28.8 a. .8 5.8 5.8
888. 888. 888. 888. 888. 888. 888. 888. 83:?
A88 88 8.. 8 A8. 8 5.8 Go. 8 G~8 88.8
888. 888 888. ~88. ~88 .888 _88. ~88 083
8.8 :88 $88 88.8 8.8 a~8 8~.8 3.8
.88. .88. 888.8 :8 _ 88 888. .88 ~88 W88 88 .8825
5.8 88.8 :88 328 A88 38.8 E8 :48
888 88 888 .88. 888.8 88 ~88 888 8E. 88
A88 2:. 8 G~. 8 E8 888 A88 A28 58
_88 ~88 88. ~88. _88 ~88 .88. ~88 8% 88885
a~8 3.8 :8. 8 5. 8 cm. 8 5.8 A-. 8 c .8
888. 888. 888. .. _88. 888. .18.? :88. .188. 8E. 50m
as; >_ 85 .88 2 858m m<< >_ m<< 28.2 >_ 8.28.2

 

3.83 .EoBP—u :3: swamp—«3 Z wmv .3.— uommmohmn: noun—:8 ammo. hot—5 +8339..." 2.: .8.— 3_=mo.. :cmmmouwmu ago—nEcU

150

 

82.8.8: 88% 8 26 038 £5 E 88.an 8: Pa 8338., .9556 £888.88 2. +

885838 E 83:88 . .«o 2...; 82038.

8. 8 88.8.8... .2. .88 a 8858: : .88. a 88585 .

 

~88 88 8.8 38 88 :8 88 88 858...-..
8.. 8... 8.. 8... 8.. 8.. 8.. 88 8882380
.28 .38 .88 .8. 8 .88 $.28 .88. 8 .88
888. 888. :88. 82.8 888 8.8 288 288 852.8
.8. 8 .88 .-.8 .88 .88 .88 .88 ...~.~8
.888 5888 .88 ~88 .88. ~88 88 :8 .88 2:8 .8 88:
.-. 8 .8. 8 .88 ...~.8 .88 .8. 8 .38 .2. 8
~88 888 .88. :88. 888 888 8.8 .28 3.28% 8:
.-. 8 .-. 8 .:..8 .88 .88 .88 .88 .88
.~.8 $.28 .88. 888 8.8 :88 8.8 :88... 0.55. .o 83520..
.8. 8 .8. 8 .8. 8 ...~.8 .8. 8 .8. 8 .8. 8 .8. 8
:88. :88. :88. :88. :88. 8.88. . :88. :88. 5.2.88 8:52
.~n.8 .8. 8 .88 .88 .-. 8 .88 .88 .88
888. 888. 888. :888. 888 888. 888 .188. 85. .588
.2. 8 2... 8 .88 .88 .88 .28 .88 .8. 8
.88 888 ~88 :88 888 888 8. .8 58 8058.53.
.88 .88 .28 .8. 8 .8. 8 .38 .88 .28
.288 :28... .88 888 888 «.88 88 :88... 888888 as;
.38 .-. 8 .88 .88 .88 .8. 8 .88 .8. 8
888. 888. 888 ~88. 888. .8. .8. «.88. ~88 82. "8:8".
.88 .88 .88 .88 .28 .38 .-.8 .88

 

151

Complete regression results for bean yields and net revenues“ survey least squares

_Egression for 436 Nicaraguan bean growers, 2003-04

 

 

 

Yields IV Yields Net revenues IV Net
revenues
Bean price 0.003 0.005 10.165*" 0.6777""'
(0.69) (0.86) (8.06) (7.27)
Insecticide price -0.087""I -0.098" -24.090""' 4.8012”
(2.46) (2.17) (2.78) (2.56)
Herbicide price -0.043 -0.029 -l.699 -0.2614
(0.81) (0.44) (0.14) (0.30)
Seed price 0.007 -0.093 -66.985" -5.3448
(0.04) (0.44) (2.3 1) (2.07)
Fertilizar price 0052‘" 0056"" -12.914”* -0.7954"”'”"
(4.20) (3.95) (3.86) (3.35)
Oxes price 0028‘" .0034" -2.868 -0.3555
(2.78) (2.20) (1.12) (1.48)
Transport cost 0.101 0.091 12.417 0.8754
(1.22) (1.01) (0.55) (0.63)
Wage 0.053 0.037 -0.888 -0.4314
(0.79) (0.47) (0.05) (0.33)
Altitude 0.01 1‘ 0.016" 2.145“ 0.2293”
(1.96) (2.23) (1.81) (2.34)
High pest level -3.262"‘" 2965"" -556.864"' 88.127”
(3.21) (2.64) (1.96) (2.14)
High disease level 2.469 4.223” 863.120" 89.117?“
(1.41) (2.01) (2.24) (3.28)
High slug level -3.1 12*" -1 .917 -778.867"" ~35.065
(2.64) (1.17) (2.66) (1.35)
Municipal distance 1232" 2.408‘I 597.151""' 59.7042?”
(1.74) (1.91) (3.05) (2.79)
FF S -0.439 -19.670 264.924 -276.0538
(0.19) (1.25) (0.43) (0.99)
Other IPM -0.577 -2.650 -191 .449 21.22
(0.56) (0.74) (0.70) (0.36)
FFS & other 1PM 0.218 -21.606” 50.015 -233.6391
(0.12) (2.35) (0.12) (1.65)
Inﬂuenced 2.068 0.980 627.566 42.602
(1.01) (0.32) (1.17) (0.87)
Years after graduation 0.1 14 8.983 -114.307 136.4294
(0.11) (1.44) (0.41) (1.25)
Adopted AAE 0.825 2.464 29.519 28.6494
(0.81) (1.35) (0.12) (0.96)
Adopted botanicals -0.104 1.861 -166.807 -7.224
(0.1 l) (1.08) (0.69) (0.25)
Adopted traps 1.145 2.434‘ 215.535 19.9775
(1.05) (1.68) (0.74) (0.95)
Observed more beneﬁcials -2.308" -0.382 -254.715 3.4447
(1.88) (0.21) (0.89) (0.14)
Apante season -3.150 ~5.709 -l,308.192* -l30.3656"'
(1.27) (1.27) (1.85) (1.71)
Postrera season -1 .228 -1.023 -308.223 -21.8674
(1.01) (0.61) (1.35) (1.10)
Female head -1.642 -2.169* 469.010 -32.3588
(1.49) (1.83) (1.47) (1.32)
Years of education 0.175 0.175 -0.558 -0.2474

 

152

 

(1.28) (1 .08) (0.02) (0.13)
Remittances -2.100"”" -2.012’ 429.778" -29.30881“‘
(2.36) (1.84) (2.00) (1.84)
Total land -0.005 -0.004 -1 .968" -0.0751
(1.37) (0.92) (2.23) (1.13)
Number of children -0.404 -0.317 ~120.343‘ -7. 1267
(1.46) (1.01) (1.86) (1.50)
Percentage of female 0.434 0.233 481.903 23.6549
(0.20) (0.10) (0.91) (0.61)
Has electricity 0.811 0.127 -45.536 -12.5814
(0.57) (0.07) (0.19) (0.61)
Heads of cattle 0.045 0.037 1 1.377 0.1428
(0.65) (0.47) (0.66) (0.1 1)
Constant 4.399 -14.383 4,538.337' -606.619"
(0.42) (0.84) (1.70) (2.21)
Observations 436 436 436 436
R—squared 0.41 0.10 0.47 0.24

 

"' Signiﬁcant at 10%; " signiﬁcant at 5%; *" signiﬁcant at 1%
Absolute value of t statistics in parentheses
+ 73 community dummy variables are not reported in this table due to space limitation.

 

153

Complete regression results for the incidence of acute health symptoms+ survey
ordered probit regr_ession for 436 Nicaraguan bean growers, 2003-04

 

 

 

Respiratory Eye irritation Stomach ache Blurred vision
difﬁculty
Bean price 0001 -0.002" 0.001 -0.001
(0.65) (1.81) (0.41) (0.87)
Insecticide price 0013* -0.003 0.013‘ 0.007
(1 .78) (0.33) (1.85) (0.89)
Herbicide price 0.005 -0.007 0.007 -0.000
(0.38) (0.94) (0.72) (0.07)
Seed price 0.068 0.009 0097"“ 0068“"
(1.61) (0.26) (2.88) (3.11)
Fertilizer price -0.003 0.002 -0.004"' -0.002
(1.11) (0.94) (1.76) (0.71)
Transport cost 0.019 0.038" -0.040"“'I 0.01 1
(0.90) (2.29) (2.54) (0.61)
Wage 0.028" 0.005 0.008 0.020‘
(2.39) (0.33) (0.50) (1.82)
Altitude -0.000 -0.000 0.000 -0.000
(0.17) (0.80) (0.39) (0.63)
Municipal distance -0.007 0.020 0041"" 0.013
(0.40) (1 .32) (2.68) (0.88)
FFS 1.137" 1.002‘ 0.156 -0.021
(2.18) (1 .89) (0.28) (0.04)
Other IPM 0.009 0.218 —0.092 0.155
(0.03) (1.15) (0.42) (0.75)
FFS & other IPM 0.196 0.017 -0.019 0.255
(0.57) (0.05) (0.05) (0.96)
Years after graduation -0.452"‘ -0.156 -0.185 0.230
(1.69) (0.64) (0.70) (0.95)
Inﬂuenced -0.341 0.636“ -0.276 -0.332
(0.84) (2.30) (0.83) (1.17)
Protective devices 0354"" -0.078 -0.236"' -0.188*
(2.79) (0.69) (1.81) (1.95)
Homemade devices 0604"" 0.234 0.267 0.101
(2.80) (1.14) (0.94) (0.63)
Hired applicator -0.017 -0.422‘ 0.184 -0.204
(0.06) (1.96) (0.75) (0.94)
Apante season 1.060 -0.091 -1 .203“ -0.260
(1 .58) (0.13) (1.92) (0.40)
Postrera season 0.377 -0.328 -0.499"' -0.l84
(1.25) (1.08) (1.93) (0.68)
Female head 0.132 0.290 -0.075 -0.222
(0.66) (1.18) (0.27) (0.89)
Years of education 0.051 0.000 -0.016 0.008
(1.44) (0.02) (0.57) (0.28)
Remittances -0.179 -0.035 -0.532" 0.310
(0.65) (0.17) (2.47) (1.54)
Total land -0.001 -0.001 0.000 0.000
(0.85) (0.89) (0.81) (0.08)
Number of children 0.074 -0.088"' 0.031 0.050
(1.43) (1.67) (0.63) (1.15)
Percentage of female 0.538 0.849" -0.578 0076
(0.91) (2.02) (1.15) (0.16)

 

 

Has electricity -0.187

(0.66)
Heads of cattle -0.006

(0.5 1)
Observations 436

0.261
(1.34)
0.027"
(2.13)

436

-o.150
(0.73)
0.001

(0.11)

436

0.063
(0.38)
0.004
(0.33)

436

 

"' Signiﬁcant at 10%; " signiﬁcant at 5%; *" signiﬁcant at 1%

Absolute value of t statistics in parentheses

+ 73 community dummy variables are not reported in this table due to space limitation.

 

155

Complete regression results for observed levels of beneficial insect populations+
survey probit and ordered probit regressions for 436 Nicaraguan bean growers,

 

 

2003-04
Adequate level of beneﬁcials Level of beneﬁcial insects
Bean price 0.001 0.000
(0.47) (0.09)
Insecticide price -0.014 0.012
(1.03) (1.03)
Herbicide price 0.031 0.023
(1.64) (1.44)
Seed price -0.056 0.126
(0.50) (1.44)
Fertilizer price -0.008"”'I 0.004
(2.21) (1 .00)
Transport cost 0.029 0.014
(1.26) (0.47)
Wage 0.048 0.047“
(1.42) (1.72)
Altitude -0.001 -0.000
(1.33) (0.1 1)
Municipal distance -0.044“ 0.018
(2.18) (0.80)
FFS 0.747 -0. 187
(0.98) (0.23)
Other IPM 0.873" 1864*"
(2.35) (4.91)
FFS & other IPM 0.820 -0.281
(0.99) (0.36)
Years aﬁer graduation 0.464 0.492
(1.35) (1.35)
Purchased protective devices -0.255 -0.066
(1.65) (0.45)
Homemade protective devices 0.073 0.713"
(0.25) (2.54)
Hired applicator 0.314 -0.431
(0.86) (1.06)
Apante season 0.987 -0.258
(0.93) (0.24)
Postrera season -0. 182 0.481
(0.37) (0.90)
Female head 0.268 0.077
(0.77) (0.23)
Years of education 0139*" 0.092“
(3.09) (2.47)
Remittances 0.282 -0.220
(0.93) (0.71)
Total land 0.004 0.013
(0.66) (1.57)
Number of children -0.024 -0.005
(0.26) (0.06)
Percentage of female -0.531 -0.323
(0.64) (0.39)
Has electricity -0.668 -0.312
(1 .55) (0.93)
Heads of cattle -0.000 -0.016

 

156

 

(0.01) (0.66)

Constant 6.558“
(2.02)
Observations 2 1 2 144

 

" Signiﬁcant at 10%; " signiﬁcant at 5%; *" signiﬁcant at 1%
Absolute value of t statistics in parentheses
+ 73 community dummy variables are not reported in this table due to space limitation.

 

157

APPENDIX A.9

Complete regression results in Essay 3

158

Predicted probability of FFS and other IPM training participation, used as
instruments in the ZSLS estimation. Probit survey regression results for 436

Nicaraggan bean growers, 2003-04+

 

 

F FS Other IPM training
Original explanatory variables
Bean price 0.002 -0.002
(1.04) (1.07)
Insecticide price 0024““ 0.001
(3.03) (0.12)
Herbicide price -0.004 0.026"
(0.41) (2.19)
Fertilizer price 0007'"
(2.62)
Oxes price 0015" 0013*"
(1.80) (2.83)
Seed price -0.041 0.031
(1.05) (1.06)
Transport cost -0.010 -0.004
(0.30) (0.19)
Wage -0.001 -0.010
(0.07) (0.71)
Altitude 0.001 -0.002"*
(1.08) (3.14)
High pest -1 450"" 0.306
(2.76) (1.10)
High disease 0.904" -1.125”‘"
(2.33) (3.01)
Municipal distance -0.002 0.005
(0.07) (0.27)
thensionist trained in IPM 2.421 3.934
(0.95) (1.61)
Extesnionist trained in FFS 6255“" 0.368
(5.26) (0.41)
Extensionist per project -1 .350 0.635
(1.38) (0.85)
Working years of NGO 0.022 -0.016
(0.30) (0.32)
Extensionist per district 4.130“ -3.260‘
(1 .85) (1.97)
Conducted comparat. experiment 4.106" 0.710
(2.18) (0.62)
Conducted other experiments -3. 133 -l .894"
(1.37) (2.09)
Observed more yields 5.546“ 4.566“
(1.90) (1.81)
Observed more revenues -0.793 -3.042
(0.51) (1.29)
Observed less revenues 3225*"
(6.18)
Observed 1es yields 2.624‘" 2121*"
(3.31) (3.27)
Soil emphasis 3.060 1.785
(0.75) (1.34)

 

159

 

Credit emphasis
Protective devices

Hired applicator

Apante season

Postrera season

Female head

Years of education
Remittances

Total land

Number of children
Female percentage

Has electricity

Heads of cattle
Redundant variables
Adopted terraces
Adopted contour
Adopted live fences
Received food ﬁom NGO
Received credit from NGO
Received trained

Constant

Observations

" Signiﬁcant at 10%; " signiﬁcant at 5%; *" signiﬁcant at 1%

Absolute value of t statistics in parentheses

+ Community dummy variables not reported due to space limitations

-4.801***
(5.24)
0.558***
(3.28)
-0.360
(1.48)
-1.566
(1.63)
-0.566
(1.21)
1.704***
(4.34)
.0.064**
(2.08)
-0.337
(1.17)
-0.006
(0.79)
-0027
(0.32)
-1.703***
(2.90)
0.405
(1.22)
0.018
(0.57)

0.098
(0.3 1)
-0.805*
(1.97)
0.755***
(2.84)
0747*"
(2.94)
0191*
(1.68)
0.308
(0.94)
-8.442**
(2.26)

436

-0073
(0.18)
0.130
(1.06)
0.036
(0.15)
0.452
(0.58)

1.106"*

(3.20)

-0.538"

(2.05)
-0.008
(0.26)

0.420"

(2.13)
-0.004
(1.52)
0.016
(0.28)
0.602
(1.29)
0.095
(0.41)

0.036”

(2.17)

0684‘"

(3.16)
-0.110
(0.44)

-0.440"

(2.33)
-0.322
(1.05)
0391*
(1.65)
0.177
(0.72)
-1.262
(0.77)

436

 

160

 

2......

 

8.... 88.... .8...... 8...? .8? 85.2.... .o 88» 002
.3... .8.... .5. .. .8. .. .8.... .8. ..
8...... .8.... 8...... 8.... 2...... 8.... .8...... .2. 8.8.88...
.3... .8.... .2... .8.... .8.... .8....
8...... .8.... .8... 8.... 2.... 8.... 8.8.8. v... 5.3 8.8.82...
.3... .8.... .8.... .8.... .8.... .8....
2.... 8...... 3...... 8...... in... 8...... 8.8.8. .2... 5... 8.8.82...
.8. .. .....N. .8. .. .2... .8. .. .8. ..
8...... .88..- 88.... .88.... .88...- 88...- 2... .28 a. m“...
.8.... .2... .8.... .8... .8. .. .3. ..
.8.... .8... 8.... .8... 8...... 8.... z... .2...
.8.... .8... .. .8. .. .8.... .R. .. .8....
.8... .8.... .8...... :88... Km... :88... m...
.8.... .3... .2... .93. .8.... .2...
8...... 8.... 8...... .88.... 8...... .88.... 85...... 38.6.82
.2... ........
8...... 2...... .32 83...... :9:
.8.... .8....
8...... 8...... .3... .8... 8...
.. m... .. n... .8.... .. n... .5... .9...
8.... 8.... 8.... 8.... 8...... 8...... 83......
.3... .8. .. .8. .. .2. .. .8.... .8....
8.... 8.... 8.... 8.... 8.... 8.... 8.3
.8. .. .R. .. .3... .8.... .8. .. .8. ..
8.... 8...... 8.... 8.... .8... .8... .8. 28.5.8
.3... .3... .8. .. .K. .. .S. .. .2...
8.... 8.... . .8.... . .8.... 8.... .8... 8.... Bow
.. ..~. .2...
.88...- .88.... 8.... 8.0.8...
.8.... .8....
:88... :88... 8.... 8.03.8:
.83 .83.
.88.... .88.... 8.... 8.8.38...
.2. .. .8. .. .8.... .8.... .R. .. .8....
8.... 88... 8.... 8...... 8...... 8...... 8.... 58..
8.0.8.... >. 8.0.8:... 86...... >. 8......8: 8.2.88. 2 8.0.88.

 

3-83 .EoBeuw :3.— naswauaﬂz omv .8 nemmmoawmmoaasvm ammo. hots... +2382. 0333....— ..8 3.58.. nommmwmwu. 3038.6

161

 

.8.... .88 .8. .. .8.... a... .. .8.~.
8.... :88... .8...... 8...... .88.... :88... 82.32 .8»... a. m...
.8.... .8.... .8.... .3. .. .8.... .8....
.8... 8...... 8.... 8.... 8.... 8...... 8.2. .28. a. m...
.8.... .3.... .8.... .8. .. .8.... .8....
8...... 8...... 8...... 8...... 8.... 88.... 8.2. .28... a m”...
.8.... .8. .. .N... .. .8.... .8.... .3....
8...... .8...? 88.... 8...... 8.... N8... 288......» .28 a. m...
.8.... .8.... .8... .8.... .8.... .8....
3.... 8.... 8.... 8...... : .... 8.... 888......» 3.8.888 a .0...
.8.... .8.... .8.... .8. .. .8.... .8....
.8... .8... .. .3... 88... 88.... .8.... .22.... .2. 888.83.... a. m...
.8.... .8.... .88. .....~. .8. .. .8....
.8.... 8...... .88.... .88.... 8...... 8.... .8...... .2. 8.8.82.8 a m...
.8.... .8.... .8. .. .8... .8. .. .8. ..
8.... 8...... .8...... 8...... 2...... 8...... 682.2...» .6 as. 002 a. m...
.8.... .8.... .8.... .8. .. .8. .. .8....
8...... 8...... 8...... 8...... 3...... 8...... 8.8.2. m... 5... 8.8.82.8 a. 8.
.8...... .8.... .8... .8... .3.... .8....
.8... a 8.... .83.... .8.... 8.... 8.... 8.8.8. 2.. 5.3 8.8.888 a. m...
.8. .. .8.. .8.... .8.... .8.... .. m. ..
.8.... .88... 8.... .8... 8.... 88... 282...... .85
.8.... .8.... .8.... .8.... .88. .8. ..
8.... .8.... 8...... 8.... .82... 8. .... 288.5 :8
.8.... 2...... .8.... .8. .. .3.... .8....
88... 8...... .8.... .3 8.... 8.... .8... 8.2.. .28..
.8.... .8. .. .8.... .8. .. .8.... .8....
2...... 8.... 8.... 8...... .8.... 8.... 882,2 2...:
.8.... .8. .. .8.... .8. .. .3 .m. .8....
8a.? {Naodn :56- omccdl EIRNNd. 3!..MNN61 mEoTA 3:»;—
.8.... .8.... .8.... .8.... .8. .. .8....
8.... 8.... 8.... .8.... 8...... 88.... 255...... 2....
.8.... .8.... .8.... .8. .. .5... .8. ..
8...... 8...... 8.... 8.... 8...? 88... 82....2..8 3.8.2.88
.2... .3... .3.... .8.... .8... .. m. ..
8...... 8.... 8.... 8...... 88.... 8...... .28.... .2. 828.82....
.8.... .88. .8. .. .8... .8.... .8....

 

162

 

5.5.5: 3...... o. 2.6 038 25 a. 08.2.0. .9. 2.. 838...... 5:63. b.5888 Q. +

38552... 5 8.8.8.... . .«o 2...... 329......

2... 8 8858.. .: .2.“ 8 82.8...m : .8... 8 .8.....8... .

 

.8.. 8... 8... 3.... 8.. 8... 8.8.8-...
.8... .8. .. .8...... .8... .8. .. .8...
8...... 8...... :88... :83... . . . .... :8... 8.880
.8.... .8... .8.... .8.... .8.... .8....
8...... 8...... 8...... 8...... 8.... 8...... 8.282. 8:
.8.... .8.... .8... .3 .. .8... 8....

.8.... 8...... .8.... .8.... 8.... 8.... 88.8.8 228..
.8... .2... .8... .3.... .8... .2...

8.... 8.... 8...... 8...? 8...... 8...... 8.2.8 8 .882
.3.... .8... .8... .8.”. .8.... .8....
86.6- cocd «.8256 8.9256 iicood iatcood 93— Each.
.8.... .8... .3.... .8.... .8. .. .8....
8...... 8...... 8...... 8...... 8...? 2...... 8.88.83.
.8. .. ..... .. .8.... .. .8. .8... .8....

8.... 8...... 8...... 8...... 8...... 8...... 8.88.. 8 88>
.3.... .8. .. .....N. .88. .8. .. .8. ..

.8.. 8.... :88... :88... .8... 8...... .82. 228...
.8... .8. .. .8.... .8. .. .8.... .3...

8.... 8.... 8.... 8.... 8...... 8.... 888 .88...
.8... .8.... .8.... .8.... .8.... .8....

.8.. 8.... 8...... 8.... .8...... .8... 888 28....
.2... .3.... .8.... .2... .2... .3...

8...... 8...... 8...... 8...... 8...... 8...... .3888 8....
.8... .2... .3. .. .8... .8.... .. m...

8.... 8...... 2...... .8... 8.... .8.. 8.888.. .8.. a. m...
.8...... .8... .8.... .88. .8... .8...
8...... 8...... ..8.. .8.....- .2... 8...... .888... =8 a m...
.8.... .8... .8. .. .8. .. E... .2...
8...... . 8...... :8... 8.... .8.... :83... 8.2. .28. a m...

 

163

164

~85 555.? :5? 5N5? 355 25.? ~85 58.? 85:55 5 as: 82
8555 8555 8555 85.3 5.55 @555 8555 G5. a
525 55.5 :55 5:5 ~25 :82 ~35 .255 .855 E .235.—
am. a 5. c $555 @555 Ed :25 a: 5 an. s
5555 :55 .25 ~55 2.255 2.555 .58: 5555 5:555 m": a? .505
55.8 85.8 $555 555 25.8 8555 $555 2555
2 _.3. $5.? 5555- :5? $5.? 525 25.2- 25.? was“: 2.: as, .28va
255 $55 Gm. : :55 8555 85. a $5. a a5. a
2.5 55.? 55...? .32.? 52.? 52.? $5.? :53? 2.: .25 a. m:
25.3 :25 3.9 @555 :55. c and 85.9 $5:
{ah—m6 {inn—6 5:50md friamvd oc—d 2.23:6 fin—NA 55352.6 25: .550
3555 G». c 8555 85. c @555 25.3 $555 $5.55
555.? .5555 5555 5.2.? 555.? 52 n5 .8. _- :5555 m":
58. a 35V 55V and £555 2 m. c 5.8 6555
£55 255 :5? t; 55.? :55 35.? :55 :52? 853.5 5325.2
:3. s 3.8 8555 :5. a 2 555 85.55 E55 55.8
2 _.? 58.? 255 :55 5555 355 R5? 5555 .22 52. .55
5.8 $55 8555 5.8 @555 2555 :55. s 35.8
555.? 555.? 555.? 555.? 555.? 555.? 55.? 58.? 33:2
555 65.8 35.8 @555 :2. s :86 8555 5.55
:5? 5555 85.? 85.? .255 2.255 855 :55 on“;
92.: c :5 :3. c 5. : €55 5.8 @555 555
555.? 85.? :55 .5555 555.? :55 :55 :55 58 22:55
555 2555 8555 8555 £555 555 a 555 @555
85.? 5555 25.? 85.? :55- :55 :5? :55 8:5 58m
:5. c 5.8 $555 :5. s c 555 5.8 $55 :5. c
85.? :5? :55 :55 :55 :55 :55 :55 3:5 02055:
3.55 9555 5555 A: . a $555 855 25.8 2555
:55 :55 85.? 555.? _555 :55 5555 :55 8E 8553.:
3555 555 :2. c 3. s 5. s :5. a 85.8 E. c
5555 555.? 555.? .5555- 555.? . :5? :5? ._ :5? 8E 58m
303055 303085—
58 >_ 555 555 >_ 558 m<< >_ 92 .552 2 52.552
visa

5.839.» :52. 5.52.582 emu be 8:59. 5:839. 8.3:? .23. 53.5» +2.: .3 no??? 2: .8.. 3.58.. 55859. 393850

£55
3555

3:336

@555
$5.:
£555
955
$5.:
_ $5-
$555
555.?
555
22.
5555
:5?
A855
58.?
555
5555
555
585
E55
:5?
$55
255
85.8
5555
5555

N367

9555
2.5.?
9555
RS.
8555
53.5
85.8

855
$5.:
255
5.:
5555
:53
332.6
2555
35.?
85.8
5.5-
$555
5::
an:
25.?
5.8
85.?
Ge:
:55
2555
$5.?
3555
E5
25.:
52.?
a»:
.5555
65.8
35.?
$5.:
.85.?
an:
525
$.55
52.?
25.8

585
$5.:
.255
35.:
“n: .c
2555
555.5
85.55
25.?
3555
.83..
9555
SE
8555
58.?
$555
3.5.?
$5.:
.855
£555
:55
A255
35.?
5.:
cONvdn
9555
5555
85.8
55?..-
@555
.55.?
85.55
:55
355
552
A5555

525
$5.:
:55
and
5.5
9:55
855
@555
525
5.:
~55-
$555
22

35V
_ :55
$5.55
.555
95.:
:55

c .55
35.?
$555
8.5
$5.:

«and.

35.:
5. _5
A5555
2.?
an:
555.?
25.:
5555
555
5555
35.8

255
$5.:
.5555
£555
:55
85.:
.38.—
8555
:5?
3555
535-
85.8
$3
@555
58.?
$555
$5.?
$5.:
.855
2 :55
:55
$555
:55
$5.:
.85.?
25.:
5555
9555
55.?
35.:
35.?
8N8
525
25.8
35.?
£555

52.?
35.8
Iimmmd
$5.3
timwwd
3.3
Sikh”—
35.3
:32 _. T
35.3
2.....OONS-
and
2.35....
an:
555.?
355V
255
85.9
EINVQN—
555
35.?
85.:
855
and
2.. and-
5.:
.255
$5.:
3.5
Gad
53:33.6.
85d
{*GOWd
95.:
.25.?
85.:

525
35.8
5555
85.:

«gwdm

c555
_ _55
8555
2.5.?
85.8
:55.
8555
$5.?
€55
58.?
a;
22-
25.8
53.:
$555
$55
@555
an?
@555
.55.?
$55
5555
555

«on.» T

9555
555.?
E555
553.
E55
535
$555

585
€5.55
5...... d N9—
2555
§§§ncNA=
85.9
3.2.53..—
85.:
53.?
$5.55
:22 _-
$3
:35.
85.:
58.?
£55
53.?
5.6
555555
8555
2.5.?
and
255
8555
*§§mwb.o-
ch:
2 m5
555
5:?
and
:«21-
85.3
:9 :2.—
5555
En?
an. a

”so: 333 a. 5.:
32:32 8:3: a. mum
m2»: .2»: a. 95
555503:
555585.55".
hmahxmdmmm
acumwxmﬁmmm
82§>Qmmm
2m<mmmﬁmmm
Smﬁkmﬁmmm
£8:qu 5520
£35.95 zom

«22» 333

.55: .23:

£2: .255
358598 350
5552330 ozﬁaaﬁoo

53.8% hem 5.85

165

:o_:S_E= 85% 3 26 038 £5 E 3:32 8: 25 mozatg 58:5: macaques ms +
88553:: E mocmzﬁm : .«o 029 82954:.
.x: 8 5565ch 3.... 5\5m :5 852.:in 3.. 2x6: :5 Esoﬁcwﬁ .

5:5 555 $5 555 :55 R5 :55 $5 835%.:
52. 59. 59. 52. 5:. 52. 59 5:. 22523.5
:25: :555: :5~. : :~5.~: :25: :-5: :25: :~_.~:
2.5 $55 5~55 :15} E5 525 $55 :355 5528
:35: :3. : :25: :5~5: :35: :3. : :~_5: :5~. :
58.? 255 5:55- 55.? $55 585 :85 ~25 55:82” a:
:_ :5: :3. : :555: :35: :2. : :3: :85: :3:
255 5: _5 5:55 5.55 ~25 :25 5m~5 55~5 053523 225:
:E: :52: :3: :3. : :~5.~: :55. : :5?: :~_.~:
18.? .15.? :85. 25.? :-5.? .55.? :53? :35? 522:0 :o 35:52
:53: :2: :E. : :52: :25: :5~5: :35: :5~.~:
:5555 5:55- 5555- :555? 5:55 555.? 5555 :85? 5.5 55.5
:55: :5~. : :85: 2 m5: :-5: :35: :35: :8:
585 5555 n _55 :55 _ 5.? 25.? $55 ~85 30555»:
:_~.: ::.~: :~:... : :3. : :55. : :3: :55: :-.~:
:55 :255 :55 5555 :55 :55 ~85 2.585 558585 23>
:-5: :35: :25: :85: :52: :55. _: :3: 5.:
:5? 25.? $5.? $5.? 5555- 35.? .35.? -~.? 32. 228:
:55: 55: E5: :555: :~m5: $5: :55: :25:
555.? 585 25.? ~55? :55 $55 52.? 585 :88... 82:8:
:25: :55: :35: :85: :55: :55. : :55: 55:
2.5.? $5.? 555.? 3.5.? 52.? 2...? 52.? :5? 833 35%
:85: :_m.~: :55.~: :55.~: :55. : :-. : :35: :35:
5.5.? :85? :.~35 a: :25 $5.? 85.? ~85 35.? 852 52::
:55. : :35: :v~5: :85: :85: :85: :55. : :55. :
.52.: 52.? ~25 .25 2~5 5255 .58.: .525 £8..qu 55205 5":
:85: :35: G55: :5~.~: :85: :55: :-5: :~m5:
«3.5 25.? 55 2.25.: 255 53.? S~.~ 525 £3580 .53. mm...
:~55: :85: :5~5: :555: :5. : :-5: :55. : :35:

166

Complete regression results for net revenue function+ survey least squares
regression results for 436 Nicaraguan bean growers, 2003-04

 

 

 

Insecticide IV insecticides
Bean price 10.24?" 11.288*"
(7.84) (5.02)
Insecticide price -28.534*" -31.117”‘"
(2.86) (2.80)
Herbicide price -4.47 1 -2.133
(0.34) (0.17)
Seed price -87.589*" 402.937"
(3.12) (2.52)
Fertilizar price -14.514"‘" -15.664*"
(4.19) (3.80)
Oxes price -3.l91 -4.752
(1.22) (1.07)
Transport cost 5.826 2.868
(0.24) (0.11)
Wage -5.394 -6.596
(0.30) (0.36) _
Altitude 1.953‘ 1.275 ‘i
(1.66) (1.01)
High pest level -537.381* -535.607
(1.97) (1.54)
High disease level 920.427" 1,298.593"
(2.34) (2.20)
Municipal distance 438.153" 673.228"
(2.00) (2.05)
FF S 1,001.366 -7,903.323
(1.24) (0.75)
Other IPM 441.840" -44.884
(2.10) (0.05)
FF S & other IPM 320.0443 658.789
(0.44) (0.54)
Extens. with IPM training 5552.444" -33,326.7OO
(1.77) (0.83)
Extens. With FFS training 1,581.089‘ 1,757.835"
(1 .81) (1.69)
Extensionists per project -1,232.401 -246.359
(1.42) (0.11)
NGO years of experience 185.125" 211.698"
(2.20) (2.01)
Extensionists per district 2,164.973 12,951.043
(1.11) (0.85)
Comparative experiments 823.772 9,791.349
(0.78) (0.69)
Other experiments -1,984.577‘ -186.822
(1.84) (0.07)
Higher yields -372.733 -36,401.585
(0.15) (0.70)
Higher revenues 1,693.727 34,597.145
(0.78) (0.69)
Lower yields -1 22.859 -567 .876
(0.30) (0.31)
Soil emphasis 1,01 1.601 -1,619.692
(0.70) (0.48)

 

167

 

Credit emphasis -3 .493 630.570
(0.01) (0.82)
FFS & Extens. with IPM 18,280.718 5,749.75]
(1.30) (0.15)
FFS & Extens. with F FS 741.935 2,870.838"
(0.35) (1.68)
FFS & NGO years of experience -327.830"" 27.346
(2.96) (0.05)
FFS & Extens. per project 7,125.919‘ -3,735.391
(1.77) (0.29)
F FS & Extens. per district -15,213.405* -2,992.549
(1.69) (0.14)
F FS & Comparative experiments 2,791.729“ -2,004.619
(1.69) (0.16)
FFS & Other experiments 1,266.050 -6,442.738
(1.11) (1.01)
FFS & Higher yields 6,913.019 26,752.] 12
(0.97) (0.62)
FFS & Higher revenues -10,161.417 -23,537.954
(1.36) (0.59)
FFS & Less revenues 613.107 743.859
(1.17) (0.75)
FFS & Less yields -l,61 1.658“ 2,571.37]
(1.81) (0.59)
FF S & Soil emphasis -512.618 7,513.954
(0.32) (0.75)
F FS & Credit emphasis 1,266.985" -2,856.929
(2.09) (0.62)
Apante season -1,274.282"‘ -1,257.783
(1 .81) (1.49)
Postrera season -361.551 ~710.473
(1.44) (1.44)
Female head -220.091 -269.071
(0.76) (0.83)
Years of education -6.095 -10.098
(0.22) (0.34)
Remittances -295.548 -356.240
(1.14) (1.25)
Total land -2.163"”'”'I -1.395*“
(2.88) (2.72)
Number of children -109.392"‘ -122.987*
(1.76) (1.82)
Female percentage 344.629 356.044
(0.65) (0.57)
Has electricity -58.374 -242.629
(0.28) (0.80)
Constant -675.639 - l ,796. 173
(0.24) (0.45)
Observations 436 436
R-squared 0.50 0.37

 

 

" Signiﬁcant at 10%; " signiﬁcant at 5%; "'" signiﬁcant at 1%
Absolute value of t statistics in parentheses

+ 73 community dummy variables are not reported in this table due to space limitation

168

   

11171111114111