The Economic Value of a Farmer Network:
Application to Pest Management in Iowa
Behzad Jeddi and Guilherme DePaula
Working Paper 24-WP 658
April 2024
Center for Agricultural and Rural Development
Iowa State University
Ames, Iowa 50011-1070
www.card.iastate.edu
Behzad Jeddi is Graduate Student, Department of Economics, Iowa State University, Ames, Iowa,
50011. E-mail: bjeddi@iastate.edu.
Guilherme DePaula is Assistant Professor, Department of Economics, Iowa State University, Ames,
Iowa, 50011. E-mail: gdepaula@iastate.edu.
This publication is available online on the CARD website: www.card.iastate.edu. Permission is
granted to reproduce this information with appropriate attribution to the author and the Center for
Agricultural and Rural Development, Iowa State University, Ames, Iowa 50011-1070.
For questions or comments about the contents of this paper, please contact Gil DePaula,
gdepaula@iastate.edu.
Iowa State University does not discriminate on the basis of race, color, age, ethnicity, religion, national origin,
pregnancy, sexual orientation, gender identity, genetic information, sex, marital status, disability, or status as
a U.S. veteran. Inquiries regarding non-discrimination policies may be directed to Office of Equal Opportunity,
3410 Beardshear Hall, 515 Morrill Road, Ames, Iowa 50011, Tel. (515) 294-7612, Hotline: (515) 294-1222, email
eooffice@iastate.edu.
e Economic Value of a Farmer Network: An
Application to Pest Management in Iowa
Behzad Jeddi
∗
Guilherme DePaula
†
April 5, 2024
Abstract
Climate change can lead to increased pest migration and more frequent outbreaks by altering
pest life cycles and habitats. Farmers facing increased temperatures or rainfall resort to more
pesticides, emphasizing the need for adaptive pest management. is article evaluates the
economic benets of farmer networks for pest management by applying an economic model
of social learning to a pilot network in Iowa. Our results show signicant variation in the
network’s eectiveness. We nd that networks are particularly valuable for farmers facing
high pest infestation risks, oering over $600 per acre in value against the impacts of extreme
heat.
Keywords: Farmer networks, Social learning, Pest management, Climate change, Iowa.
JEL Codes: Q12, Q15, Q16, Q54, Q55
∗
Graduate Student, Department of Economics, Iowa State University. [email protected]
†
Assistant Professor, Department of Economics and Center for Agricultural and Rural Development (CARD),
Iowa State University. [email protected]
1 Introduction
Climate change is increasingly recognized as a critical driver of pest migration, range expansion,
and more frequent outbreaks, primarily by altering their life cycles and habitats (Hall et al. (2002);
Macdonald et al. (2005); Gutierrez et al. (2008); Jackson et al. (2011); Noyes et al. (2009); Miraglia
et al. (2009)). is shi, which tends to favor pests over crops, is aributed to climate-induced
changes in the environment (M
¨
uller et al. (2010), Roos et al. (2011)). Research indicates that,
while insects can thrive in various climates, they tend to appear earlier and become more active
in warmer conditions, a phenomenon exacerbated by climate change (Rosenzweig et al. (2001);
Bloomeld et al. (2006) and Jackson et al. (2011)). As a consequence, farmers in regions experienc-
ing notable increases in temperature or precipitation are oen compelled to use higher pesticide
dosages to protect their crops, highlighting the need for adaptive strategies in agricultural pest
management.
Agricultural economists recognize the importance of farmer coordination as a key strategy
for ecient pest management. e works of Lazarus and Dixon (1984) and Vreysen et al. (2007)
underline the ineectiveness of isolated farm-level pest control eorts, particularly given the
mobility of pests. Research by Singerman et al. (2017) and Lence and Singerman (2023) further
supports the need for coordinated action over broader areas to combat mobile insect pests e-
ciently. Such collective strategies can help reduce the frequent and widespread use of the same
pesticides, a signicant factor in developing pest resistance. Hurley and Sun (2019) argue for the
establishment of farmers’ networks to promote collaborative pest management eorts across the
United States, emphasizing the role of social learning in enhancing these eorts as supported by
studies like Miranowski (2016) and Feder and Savastano (2006).
Although the benets of farmer networks for pest management and the spread of agricul-
tural technologies are well acknowledged, the development and expansion of these networks face
several obstacles. Technological barriers, particularly telecommunications challenges in rural ar-
eas, pose one part of the problem. Privacy concerns related to sharing farm-specic information
on crop and pest management also deter participation. Economic questions further add complex-
ity: determining the economic value of network participation, identifying which farmers would
gain the most, understanding the investment needed in communications and pest management
technology, and assessing whether these investments would yield protable returns are all crucial
considerations that need addressing to facilitate the growth of farmer networks.
In this article, we aim to explore the economic value of participating in networks for pest
management. To achieve this, we adapt a social learning economic model to the context of pest
management, drawing from foundational works by Foster (1995), Conley and Udry (2010), Udry
(2010), Krishnan and Patnam (2014), and BenYishay and Mobarak (2019), with a specic focus on
modeling the optimal timing for pesticide application—a critical decision for farmers. Applying
pesticides too early can incur unnecessary costs without signicant benets to production and
1
protability, while late applications can drastically reduce crop yields. Building on this frame-
work, we conduct Monte Carlo simulations of the social learning model to evaluate the economic
value of the network under standard climate conditions and assess its adaptation value in sce-
narios of extreme heat resulting from climate change.
Our simulations center on a pilot farmer network in Iowa (SIRAC), consisting of 121 Iowa
Soybean Association (ISA) farmers. is network was established to evaluate new technolog-
ical advancements in pest management and telecommunications, providing a practical seing
to assess the potential benets and challenges of integrating these innovations in a real-world
agricultural context. In our simulations, we distinguish between the contributions of three dis-
tinct learning channels: the impact of previous experience in pest management or guidance from
external sources, the practical knowledge gained through direct observation or ”learning by do-
ing” via scouting technologies, and social learning facilitated by the exchange of information
with peer farmers within the network. is approach enables us to evaluate the network’s value
across various scouting technologies and degrees of farmer experience.
We have three main results from our simulations of the farmer network.
First, our analysis indicates that the economic benets of joining a farmer network like
SIRAC vary widely under normal climate conditions, ranging from minimal gains to substantial
increases in prot per acre. Farms with advanced scouting technologies and less vulnerability to
pests—oen due to geographical isolation or lower pest pressure—tend to gain the least, while
those facing higher risks from environmental factors, market conditions, and possessing less pre-
cise scouting methods benet the most. Specically, gains vary from as low as $85 to as high as
$501 per acre, with the most signicant benets accruing to farmers who are closer to network
peers and thus receive more accurate information to manage pest infestations eectively.
Second, we evaluate the economic benets of expanding the network. Upon expanding the
SIRAC network by adding ve neighboring farms within a 30-mile radius of each existing farm,
thereby increasing the network to 605 farms, we observe the economic benets of such expan-
sion are modest without implementing targeted signal selection. However, introducing signal
selection based on geographic proximity signicantly enhances the network’s eectiveness, re-
ducing the average distance for received signals by more than 90%. is targeted approach to
signal dissemination leads to notable improvements in the economic benets of network partic-
ipation. In the lowest quantile, gains rise to $148 per acre, a 59% increase, while in the highest
quantile, gains reach up to $688 per acre, reecting a 28% enhancement. is demonstrates signal
selection’s critical role in maximizing network expansion’s value for farmers.
ird, we explore the network’s adaptation value in scenarios of extreme heat resulting
from climate change. We discover that the network oers considerable benets for farmers most
vulnerable to pest infestations, with its value in countering the eects of extreme heat on pest
infestations surpassing $900 per acre. Additionally, the network’s potential as an early warning
2
system for pest infestations could complement agricultural insurance policies, particularly as
climate change prompts warmer growing seasons. By alleviating the most severe impacts, the
network has the potential to diminish potential losses and, as a result, lower insurance premiums.
is article adds to the expanding body of agricultural and development economics research
focused on the value of social learning among farmers. Conley and Udry (2010) demonstrate
that farmers rely extensively on insights from their peers. ey use an interconnected network
of information to assess the comparative protability of varying fertilizer usage across dierent
weather and soil conditions, nding social learning nearly as impactful as personal experience in
agricultural decision-making. Udry (2010) highlights the critical role of social learning in devel-
oping extension programs. Further studies by Bandiera and Rasul (2006), Maertens and Barre
(2013), Vasilaky and Leonard (2018), Crane-Droesch (2018), Takahashi et al. (2019), Di Falco et al.
(2020), Beaman et al. (2021), and Adjognon et al. (2022) emphasize the inuence of social networks
on technology adoption, pointing out the signicant role of information sharing in enhancing
yields.
1
Despite the sparse literature on social learning in pest management, existing research under-
scores the necessity of coordinated approaches for eective pest control, highlighting the limita-
tions of isolated farm-level treatments due to pest mobility. Studies by Lazarus and Dixon (1984),
Vreysen et al. (2007), Singerman et al. (2017), and Lence and Singerman (2023) stress the impor-
tance of broad-scale coordinated treatment to address mobile pest issues, notably reducing the
overuse of pesticides and the risk of resistance
2
and Hurley and Sun (2019), underscore the im-
portance of learning from social networks.. Our study builds upon these insights by investigating
the benets of learning optimal pesticide timing through farmer networks, a central aspect given
the signicant impact of timing on pest control ecacy and farm protability.
Also, in the context of pest management, the role of farmer networks and the sharing of
knowledge among peers are increasingly recognized for their potential to reduce uncertainty and
encourage the development of innovative, pesticide-free agricultural systems (Wang et al. (2023)).
Foley et al. (2011) emphasize that a key strategy for achieving ”sustainable de-intensication” is
minimizing environmentally detrimental inputs. With these resources becoming scarcer, there
is a pressing need to enhance production eciency using equal or fewer resources, highlighting
the importance of improved resource use eciency for global food security.
3
Rebaudo and
1
Additionally, Beaman (2019) show how social connections can aect labor market outcomes, hinting at similar
eects in agricultural productivity.Munshi (2004) delves into social learning within diverse populations, highlighting
its role in the diusion of technology during the Indian Green Revolution. Banerjee et al. (2013) explored the spread
of micronance, underscoring the importance of social ties in disseminating information and innovation among
farmers. BenYishay and Mobarak (2019) illustrate that farmer networks could surpass the ecacy of traditional
extension programs at a lower cost. Krishnan and Patnam (2014) nd that the eects of social learning, especially
concerning the adoption of improved seeds and fertilizers, are more pronounced than learning from extension agents,
reinforcing the signicance of peer-to-peer learning in agriculture.
2
Other studies, such as Miranowski (2016); Feder and Savastano (2006)
3
Empirical research indicates that information disparities among farmers can result in either the overuse or
underuse of pesticides, with signicant implications for both protability and production eciency. Studies by
Babcock et al. (1992), Antle and Pingali (1994), Carpentier and Weaver (1997), Zhengfei et al. (2006), and Grovermann
et al. (2013) highlight how information gaps can lead farmers to overapply pesticides to maximize prots while oen
disregarding the environmental and health costs. Conversely, a lack of information can also lead to the underuse of
3
Dangles (2011) illustrate that farmer-to-farmer learning signicantly reduces pest infestations at
the community level, suggesting that social learning can lead to sustainable benets over the long
term. is exploration into the value of farmer networks in pest management lls a gap in the
current research and provides practical insights into enhancing agricultural practices through
improved coordination and social learning.
is article is structured as follows. Section 2 oers background information on the timing
of pesticide application and details the SIRAC network. Section 3 outlines the economic model of
social learning, adapted specically for pest management challenges. In Section 4, we detail our
simulation design, while Section 5 showcases the results of these simulations. Section 6 concludes
and summarizes the policy implications derived from our ndings. Appendix A provides in-depth
derivations of the economic model of social learning. Appendix B focuses on the most prevalent
pests in corn production. Appendix C elaborates on the simulation methodology step by step.
Lastly, Appendix D contains supplementary results.
2 Pest Management and Farmer Networks
e Timing of Pesticide Application
In corn production, predominant pest threats include Diabrotica virgifera (Western Corn Root-
worm), Diabrotica barberi (Northern Corn Rootworm), Helicoverpa zea (Corn Earworm), Stria-
costa albicosta (Western Bean Cutworm), and Ostrinia nubilalis (European Corn Borer, ECB), as
elaborated in Appendix B. Despite the distinct phenologies and environmental adaptabilities of
these pests, agricultural extension services, such as the Ohio State University Extension
4
, have
delineated three principal pest management strategies: the cultivation of transgenic maize vari-
eties expressing Bacillus thuringiensis (Bt) toxins; the application of insecticidal seed treatments;
and, the deployment of soil or foliar insecticides.
e recent trend towards preemptive pest management strategies, particularly adopting Bt
maize, marks a proactive approach to controlling pests. While Bt maize has signicantly re-
duced pest-related damage, the Bt bacteria are eective against only certain pests, and growing
Bt corn requires establishing non-Bt refuge areas to prevent pests from developing resistance.
Furthermore, the appearance of Bt-resistant pests in some species highlights the limitations and
challenges of current pest management methods, calling for additional suppressive tactics to
maintain agricultural productivity. According to the United States Department of Agriculture
National Agricultural Statistics Service, pesticide applications were the leading method of pest
suppression in U.S. corn production in 2021, as reported by 43% of respondents. Additionally, the
most common practice for monitoring was the use of weather data to time pesticide applications,
utilized by 60% of respondents NASS (2014).
pesticides, as shown by Carrasco-Tauber and Mo (1992), Chambers and Lichtenberg (1994), Fernandez-Cornejo
et al. (1998), and Lansink and Carpentier (2001), resulting in production ineciencies and potentially lower yields.
4
Ronald B. Hammond and Eisley (2014)
4
e timing of pesticide application is critical in eective pest management. Delayed appli-
cation risks escalating pest populations beyond control, while premature treatment may result in
ineectiveness against population growth, necessitating further, costly interventions. Moreover,
pinpointing the optimal timing for pesticide deployment is complex, inuenced by variables such
as climatic conditions and farm management strategies, including crop rotation, eld congura-
tion, and seed selection. To navigate these challenges, farmers employ a variety of methods to
determine the most eective timing for pesticide use. Predominantly, this involves scouting for
pests and utilizing thermal summation techniques, such as growing degree days, to forecast pest
population densities and determine the ideal timing for pesticide application.
For example, in their guidance on managing the ECB, Hodgson and Rice (2017) from the Iowa
University Extension Services emphasize the ephemeral ecacy of insecticides, underscoring the
necessity for timely application. Specically, they state:
”Insecticides exert their lethality on larvae within a relatively brief window; hence, their appli-
cation must precede the completion of egg deposition. Postponing treatment risks allowing larvae
from initially laid eggs to inltrate the plant, rendering them impervious to control measures. e
precision of application timing emerges as a pivotal factor in the successful mitigation of corn borer
infestations via insecticides.”
is guidance highlights the intricate balance required in pest management, particularly the
critical importance of synchronizing insecticide application with the pest’s life cycle to maximize
ecacy and minimize crop damage.
Examples of pests that can be managed using pesticide applications are Corn Rootworm,
Corn Earworm, Western Bean Cutworm, and the ECB Christian H. Krupke (2017).
e SIRAC Farmer Network in Iowa
Agronomists, engineers, and economists from Iowa State University, Missouri Institute of Tech-
nology, the University of Kentucky, and the Iowa Soybean Association (ISA) (reference here)
are designing and testing the Smart Integrated Farm Network for Rural Agricultural Commu-
nities (SIRAC), which is a connected farm network in Iowa
5
. SIRAC’s goal is to facilitate data
sharing, knowledge exchange, and coordinated responses to production threats, contributing to
community-led decisions on biological pest spread and mitigation.
Figure 1 illustrates a simulated depiction of the SIRAC network across Iowa, with a total of
121 farms. is simulated network was built using the actual pairwise distance between farms
provided by ISA, to protect the condentiality of each farm’s precise location. Starting from a
central reference point in Ames, Iowa, the simulation estimates the spatial positioning of individ-
ual farms, utilizing the provided distance metrics. e range of pairwise distances in the dataset is
5
SIRAC website: hps://sirac.agron.iastate.edu/.
5
notably broad, starting from less than two feet at its minimum and extending up to approximately
278 miles at its maximum.
Importantly, Figure 1 categorizes the precision of pest detection technologies as either ”low”
or ”high.” is classication hinges on the number of traps deployed across the elds, varying
from a single trap to a total of seven. Green circles are farms that have installed more than four
traps in their elds, categorized under ”high” precision, indicating an elevated level of accuracy in
pest detection capabilities. On the other hand, red squares are farms with fewer than four traps,
indicating a less precise approach to detecting pests.
6
(a) SIRAC Network
(b) Expanded Network
Figure 1: Farmers’ Network: Panel (a) displays the SIRAC Network with 121 farmers located in
Iowa; Panel (b) shows our hypothetical expanded network with 605 farmers.
To explore the advantages of a broader network, we develop a larger simulation of the
farmer’s network, expanding it to include 605 farms. We achieve this expansion by adding ve
6
e choice to use four traps as the threshold between the two groups is based on the actual number of traps
installed per farm in the SIRAC network, which suggests that farms with more traps can beer understand when
pests are becoming a problem, allowing them to use pesticides at the right time. is can help prevent pest damage
more eectively than farms with fewer traps.
6
additional neighboring farms at various distances to each farm in the original network. We ran-
domly select these new neighbors within a 30-mile radius, with distances ranging from about
0-27 miles and an average distance of around ve miles.
Figure 1 shows this expanded network in Iowa. In this depiction, farms are dierentiated
by their pest detection capabilities using black squares and purple shapes. Black squares indicate
farms that have a basic level of pest detection technology, while purple shapes denote farms with
a more advanced or precise system for detecting pests. We base this distinction on the number of
traps a farm has installed—we consider farms with more traps as having beer quality information
about when to treat pests.
By increasing the number of traps, we expect farms to have more accurate data, helping
them decide the best time for pest treatment. is expanded network model aims to show how
a larger, more connected community of farms can enhance pest management through improved
detection and information sharing.
3 A Model of Learning about Pesticide Application
We adapt the farmer learning process concerning pesticide application by extending the target-
input model. Development and agricultural economists have widely embraced this economic
model, originally developed by Foster (1995) and Jovanovic and Nyarko (1995), due to its sim-
plicity and adaptability in modeling learning across various farming inputs, as highlighted in
studies by Beaman et al. (2021), Conley and Christopher (2001), Vasilaky and Leonard (2018),
and Ariel BenYishay (2019). Our adaptation of the target-input model focuses on identifying the
optimal timing for pesticide application as the key uncertain input requiring farmer education.
Essentially, the target-input framework conceptualizes learning as a reduction in the variance
associated with production inputs. For instance, a farmer initially inexperienced in pest manage-
ment might face signicant uncertainty in determining the optimal pesticide application timing,
reected in a high variance of estimates. However, as the farmer’s experience and knowledge
expand, this variance is expected to diminish, implying an improvement in precision and farmer
protability.
We denote the optimal timing for pesticide application for a given farm i in season t as
˜τ
it
, which we decompose into two components: (a) the universal optimal timing across farms,
represented by τ
∗
; and, (b) a farm and season-specic term, µ
it
. We posit that µ
it
behaves as an
independently and identically distributed (i.i.d.) normal random variable, characterized by a mean
of zero and a variance denoted by ϑ
2
µ
. is formulation allows us to capture the commonality in
optimal pesticide application timing across dierent farms and the unique variability each farm
and season might introduce.
˜τ
it
= τ
∗
+ µ
it
(1)
Extension agencies and pesticide suppliers oer guidance on τ
∗
, the recommended timing for
7
pesticide application within a specic region. However, µ
it
, which accounts for farm-specic or
within-farm variations due to dierences in climate, vegetation, and management practices, can
vary signicantly. e precise value of µ
it
for any given season is not known to the farmer. In-
stead, farmers typically rely on their personal experience with pest control to make an informed
estimate of µ
it
. To accommodate the inherent uncertainty in estimating µ
it
, we model it as an i.i.d.
normal random variable with a mean of zero and a variance denoted by ϑ
2
µ
. is approach ac-
knowledges the unpredictable nature of agricultural conditions and the adaptive strategies farm-
ers employ in pest management.
Pest Population: Farmers determine the optimal time for pesticide application by assessing
the pest population on their farms. is process of determining the best application timing is
essentially a learning exercise about the pest population dynamics. To this end, we incorporate a
straightforward model of pest population growth into the target-input framework, aligning with
established pest management research that posits pest populations expand exponentially (add
references about pest population growth models). e equation for the pest population P
w
at
farm i, w weeks into the season, is given by:
P
t
= P
0
e
rGDD
(2)
where P
0
represents the initial pest population at the start of the season and r denotes the pest’s
internal growth rate. To address the pest population’s uncertainty, we dene P
t
as a random
variable normally distributed with mean µ
P
t
and variance ϑ
2
P
t
. As farmers scout their elds,
they accrue more precise information about µ
P
t
, allowing for an increasingly accurate estima-
tion of future pest populations and, consequently, a reduction in the variance of these estimates
throughout the season. To simplify the notation, we exclude the season subscript in subsequent
equations.
Production Function: At the heart of the target-input model lies a production function,
which models the farm’s maximum potential yield and incorporates a loss function. is loss
function quanties yield reductions aributable to deviations from optimal input utilization.
Specically, for pest management, we dene the loss function in relation to deviations from the
ideal timing of pesticide application:
q
i
(τ
i
) = ¯q
i
− αg(GDD)(τ
i
− ˜τ
i
)
2
(3)
where q
i
represents the quantity of corn produced per acre on farm i, with ¯q
i
denoting the maxi-
mum potential yield. Yield losses occur as a result of deviations from the optimal pesticide appli-
cation timing, ˜τ
i
. To streamline the model, we exclude other production inputs such as labor and
fertilizer, although their inclusion would not alter the target-input model’s outcomes. e pest
growth function g(GDD) is determined by growing degree days (GDD), reecting temperature’s
role in pest development. We expect yield losses to escalate as the season progresses, correlating
8
with increased pest density in the eld. e parameter α serves as a scaling factor that measures
the impact of timing deviations on corn production sensitivity.
e quadratic formulation of the production function serves as an approximation for the loss
function surrounding the optimal timing of pesticide application. is symmetrical loss structure
around the optimum reects the potential for yield reductions due to premature applications
and losses stemming from infestations caused by delayed applications. As deviations from the
optimal timing expand, we expect the yield losses associated with signicantly late applications
to escalate. is quadratic approach is critical within the target-input model framework, as it
facilitates the analysis of farmer protability in relation to the variances of input levels.
Farmer Protability: To understand how variances in the timing of pesticide application
inuence a farmer’s expected prots, we analyze the farmer prot maximization problem. A
farmer, denoted as i, selects the timing of pesticide application, ˜τ
i
, and the number of scouting
trips, S, aiming to maximize her expected prots E(π
i
):
E(π
i
) = max
τ
i
,s
pq
i
(τ
i
) − s
i
× a
i
× r − c
i
(4)
where p represents the price of corn; r is the average scouting cost per acre; a
i
denotes the
total acreage scouted; and c
i
captures the total cost of fertilizer application. Maximization of the
farmer’s expected prot implies that τ
i
= E(˜τ
i
) = τ
∗
, suggesting the farmer will opt for the
average optimal timing. e optimized prot function then becomes:
E(π
∗
i
) = p ¯q
i
− αg(GDD)(ϑ
2
˜τ
i
+ ϑ
2
µ
) − s
i
× a
i
× r − c
i
(5)
is equation links expected protability directly to two types of variance. e rst, ϑ
2
˜τ
i
, captures
the uncertainty around the optimal timing of pesticide application, which farmers can reduce
through learning from their own past experience and through their interactions within their so-
cial network. e second, ϑ
2
µ
, represents uncontrollable random eects, like specic weather
events or unique pest developments, that learning processes cannot mitigate. Equation 5 is a
central result of the target-input model, connecting the learning mechanisms directly to farmer
protability. Using Bayes’ rule for a normal distribution allows for the derivation of a straight-
forward equation for the variance of the uncertain input choice, ϑ
2
˜τ
i
, highlighting the impact of
both experiential learning and social learning on decision-making processes.
Learning by Doing: Farmers gain insights from their own experiences. We divide the grow-
ing season into weeks, starting with the farmer’s initial estimate of pest population growth based
on prior experience, weather forecasts, and chosen management practices. As the season pro-
gresses, the farmer can update her pest population estimates weekly through scouting.
By applying Bayesian updating to the population growth process, we derive an expression
for ϑ
2
˜τ
i
that incorporates learning by doing. We use the relationship between the optimal tim-
9
ing of pesticide application and the pest population to derive the variance ϑ
2
˜τ
i
conditional on an
observation of the population P
w
in week w.
7
rough Bayesian rule, we nd:
ϑ
2
˜τ
i
=
1
ρ
0
+ γ × ρ
S
(6)
where ρ
0
represents the precision of the initial estimate of optimal pesticide application timing
at the season’s start. Precision, the inverse of variance (ρ
0
==
1
ϑ
2
0
), improves with more accurate
initial estimates. For instance, experienced farmers are likely to have more accurate application
timing estimates, leading to lower ϑ
2
0
and higher ρ
0
, thus reducing ϑ
2
˜τ
i
and enhancing protability
as shown in equation 5.
e second term, ρ
S
, reects the precision of the farmer’s learning technology, such as the
accuracy of information obtained from scouting, inversely related to its variance (ϑ
2
S
). High-
quality scouting increases ρ
S
. Investment in technologies like cameras and trapping devices can
also increase the precision ρ
S
. Finally, the precision of the learning technology is multiplied
by a factor γ in equation 6 that adjusts for pest population growth characteristics and scouting
frequency, inuenced by factors such as pest growth and death rates, the number of scouting
reports, and the correlation between pest population and optimal pesticide timing. Appendix A
derives γ, which increases with more frequent scouting, illustrating a balance between scouting
frequency and technology quality. Farmers with less precise technology may need more frequent
scouting to achieve the protability levels of those with advanced technology.
8
Learning from Others (Social Learning): Farmers also benet from the knowledge and
experiences of their peers within their social networks. For instance, a farmer equipped with
advanced pest detection technology might share valuable insights about unusual pest develop-
ments with neighboring farmers, enhancing the network’s collective understanding of pest man-
agement. is exchange of information, or ’signals’, particularly regarding the optimal timing for
pesticide application, forms a critical component of social learning in pest management. Speci-
cally, in the pest management application, we dene a signal as a neighboring farmer’s estimate
of their optimal time of pesticide application. A farmer might receive N signals from peers each
season, with the quality of these signals varying signicantly.
In the target-input model, precision quanties the informational value of a signal, ρ
N
, de-
ned as the inverse of the variance of the optimal timing of pesticide application from the signal’s
sender, ρ
N
=
1
ϑ
2
ξ
+ϑ
2
τ
j
, where and ξ represents an additional error term to account for the signal’s
noise. While we initially assume signal precision from peer farmers is uniform for simplicity, our
simulations introduce variability in signal precision based on the geographic proximity among
farmers, aligning with methodologies commonly employed in learning literature (Conley 2001;
7
Appendix A details the derivation, resulting in the optimal timing’s variance conditioned on P
w
as ˜τ
i|P
w
∼
N(τ
⋆
, ϑ
2
µ
(1−ρ
τ,p
)), where ρ
τ,p
denotes the correlation between the optimal application timing ˜τ and pest population
P
w
.
8
We derive in Appendix A an equation for γ using the Bayes rule. γ =
N
[1−
e
2rGDD
e
2r
+1
ρ
2
τ |P
0
]
. γ is always positive
given that the correlation between pest population and the timing of optimal pesticide application, ρ
2
τ |P
0
, is positive.
10
Conley and Udry 2010).
9
As part of a network, receiving N signals allows a farmer to rene her estimates for the
optimal pesticide application timing on her farm. For example, learning about a peer’s observa-
tion of unexpected pest population growth could prompt a farmer to adjust her own estimates
accordingly. rough Bayesian updating, we derive a revised equation for ϑ
2
˜τ
i
that incorporates
social learning:
ϑ
2
˜τ
i
=
1
ρ
0
+ γ × ρ
S
+ N × ρ
N
(7)
is equation extends equation 6 by adding a third term, N × ρ
N
, to the denominator, reect-
ing the impact of social learning. e eectiveness of learning increases with the receipt of a
greater number of high-precision signals (N), and with precise signals from the network, high
ρ
N
. is improvement in learning reduces the variance ϑ
2
˜τ
i
, subsequently boosting farmer prots.
Moreover, the product N × ρ
N
suggests a trade-o between the quantity and quality of signals,
indicating that receiving numerous high-quality signals can signicantly enhance a farmer’s un-
derstanding and management of pest populations.
The Value of Social Learning for Pest Management: e value of social learning in pest
management is quantied by the additional expected prot a farmer gains by integrating infor-
mation from peers into her decision-making process regarding the uncertain timing of pesticide
application. e prot function in equation 5 denes this concept, which translates the impact of
learning into monetary terms. As learning progresses, the variance ϑ
2
˜τ
i
diminishes, leading to an
increase in expected prot. erefore, the value of social learning is represented by the dierence
in expected prots—with and without the inuence of peer learning, as detailed in equation 7. To
quantify this value, ∆, we calculate the dierence between the expected prot function incorpo-
rating social learning (via equation 7) and the expected prot absent social learning (via equation
6)
10
:
∆ = E(π
∗
i
|with social learning) − E(π
∗
i
|without social learning)
= −αg(GDD) [
1
ρ
0
+ γ × ρ
S
+ N × ρ
N
−
1
ρ
0
+ γ × ρ
S
]
(8)
We use equation 8 to simulate the value of social learning pest management within a network of
farmers in Iowa. Naturally, farmers can learn more from their peers than about optimizing pesti-
cide application. us, our simulated values for social learning will underestimate the total value
of learning within the network. However, we can extend the framework to other applications
with alternative farming inputs such as fertilizer and labor. e value of learning will be higher,
9
We also assume that signals from dierent farmers within a network are independent when applying Bayes
updating. More specically, we assume that cov(µ
i
, µ
j
) = 0 for any pair of farmers (i, j). e independence
assumption would likely be violated in a large network of close farms or elds where signals of eld i and eld j
received in the same week would be correlated. However, in networks such as the SIRAC network in our simulation,
the minimum distance among farmers is less than 1 km. We could extend the model to incorporate correlation among
signals explicitly but we leave this extension for future work.
10
Note that the learning eect dened in terms of expected prots captures a reduction in losses because of
learning (see equation 5. e gain in protability from social learning is therefore the negative of the dierence in
losses.
11
the larger the uncertainty about the optimal use of an input or the optimal choice of management
practice. e application to pest management is important because of the uncertainty about the
key choice of the time of pesticide application. Furthermore, we can extend the framework for
the more general case of multiple pests.
4 Methods: Monte Carlo Simulations
To evaluate the economic value of a network of farmers engaged in pest management, this study
employs Monte Carlo simulations to project the expected prots of farmers, with and without the
eects of social learning. ese simulations involve generating thousands of random parameter
samples from the economic model for the network’s value (as in equation 8). ese samples are
the basis for calculating the expected gain and the distribution of economic gains aributable to
the farmer network.
In each simulation, we estimate the farmer’s expected gain under three distinct scenarios
that represent dierent learning mechanisms: previous knowledge; scouting (learning by doing);
and, social learning. e initial scenario, termed the baseline model, assumes farmers have no
access to external information to determine the optimal timing for pest control, relying instead
on their knowledge and previous experiences. As a result, in this scenario, farmers’ expected
losses are the highest due to a discrepancy between their chosen timing for pest management
and the ideal, most eective timing. Next, we assess the impact of learning through scouting,
which involves direct experience in the eld. Finally, we explore the benets of incorporating
social learning within the farmer network. With each addition of new learning channels, we
calculate the decrease in losses aributed to pest infestations.
e simulation of the distribution of a farmer’s expected gain involves drawing a sample of
10,000 observations from the distribution of the model parameters. Our model is based on two
primary sets of parameters, summarized in table 1. e rst includes endogenous parameters,
which are inuenced by the farmer’s decisions. ese include the frequency of scouting activ-
ities, which reect a farmer’s eort to monitor pest infestation levels. For the purpose of our
simulations, we assume that farmers conduct scouting weekly throughout the farming season.
Another key endogenous parameter is the farmer’s initial estimate regarding the optimal timing
for pesticide application. We categorize initial knowledge into two levels: low initial precision,
representing farmers with limited experience and knowledge about the optimal timing for pest
treatment; and, high initial precision, indicative of farmers with extensive experience. In our
simulations, due to the absence of specic data regarding farmers’ knowledge and experience,
we make the assumption that all farmers within the network possess limited knowledge and
experience. is assumption does not impact the calculation of the network value because we
dierence out the value of initial experience.
e simulation also incorporates two additional endogenous parameters: the precision of
the pest detection technology; and, the precision of the informational signal from the farmer net-
12
Table 1: Simulation Parameters
Parameter Values
Number of farmers- SIRAC network 121
Number of farmers- expanded network 605
Initial precision of farmer’s estimate Uniform(0, 1)
Precision of the scouting Number of traps / Maximum number of traps
Precision of the signals
1
10
P
10
i=1
signal
i
× w
i
Distance from signal sender: Signal weight (w
i
)
0 to 10 miles 1.00
10 to 25 miles 0.75
25 to 50 miles 0.50
More than 50 miles 0.25
Pest growth rate (g) determined by
dP
t
dt
= rP
t
(1 −
P
t
κ
)
Pest death rate (δ) Uniform(0, 0.4)
Pest carrying capacity (κ) N (22, 0.5)
Pest intrinsic growth rate (r) 1 + g − δ
Pest initial population (P
0
) N (2, 0.5)
Corn Price N (6.4, 0.83)
Corn yield N (173, 5)
Frequency of scouting 10 per season
antity of the received signals 10
Growing Degree days (GDD) N (1500, 500)
Number of simulations 10,000
Note: We generated 10,000 observations with replacement for the parameters following a normal distri-
bution. e precision of the signals was assessed based on the quantity of traps at the sources of these
signals. To incorporate the aspect of trust that farmers aribute to these signals, we allocated weights
based on the spatial distances between the signal origins and the recipients.
work. We quantify the accuracy of the pest detection technology based on the number of traps
a farmer installs in their eld, which we calculate by dividing the number of traps placed in a
given eld by the maximum number of traps used, which is seven. Consequently, the precision
parameter for the technology varies from 0.14 to 1. A precision value of 0.14 indicates a eld
with just one trap installed, suggesting minimal technology deployment. Conversely, a precision
value of 1 denotes the installation of seven traps, the maximum considered in our study, indi-
cating the highest level of technological deployment for pest detection. In our simulations, we
categorize scouting precision as low or high. Low precision scouting corresponds to the average
precision for farms with trap counts at or below the network’s median (four traps). We determine
high-precision scouting by the average precision of farms with trap counts above the network’s
median.
To quantify the precision of the informational signal received from the farmer network, we
employ a proxy combining two elements: the count of traps installed in the originating eld; and,
the spatial distance between the signal senders and the recipient. e number of traps installed
measures the information accuracy shared by the senders, with a lower count indicating reduced
precision. Additionally, we compute a weighted average for the signal’s precision as received by
a farmer, where the distances from the senders to the receiver determine the weights. For each
13
farm, we cluster the neighbors based on distance and assign weights for each group of neighbors.
In our simulations, we account for the distance between signal origins and destinations by
assigning weights to capture a farmer’s trust in the signal’s relevance. Each farmer receives ten
signals, with the weight of each signal determined by the distance from its source. Specically, we
weight signals originating from within a 10-mile radius at 1, acknowledging the strong potential
for social ties or trust among nearby farmers. We assign signals from 10 to 25 miles away a weight
of 0.75, those from 25 to 50 miles receive a weight of 0.50, and we give signals from sources over 50
miles away a weight of 0.25. is system reects the understanding that farmers are more likely to
observe and trust their close neighbors’ farming decisions and outcomes. Our simulations assume
that each farmer receives ten signals. We compute the weighted average of the ten signals for
each signal-receiving farmer using the assigned weights. High precision refers to the average
of signals with precision greater than the median value for all signal receivers. Conversely, low
precision signal refers to the average of signals with precision less than the median value.
11
e second set of parameters are exogenous factors, which are external to the farmer’s con-
trol and stem from the broader environmental context. A key exogenous parameter is the growth
and death rate of the pest or insect population, as it directly aects the population size and dy-
namics of the pests over time. e growth and mortality rates among the pests are contingent
upon various factors, one of which is the total GDDs accumulated during each season. GDDs are
a measure of heat accumulation over time and serve as an important gauge for understanding the
development and reproductive cycles of the pest population. Additionally, the carrying capacity
for pests, indicating the highest pest population that the agricultural ecosystem can sustain, is
another important exogenous parameter. Multiple ecological variables, such as the availability of
host plants, the presence of natural predators, and the general environmental conditions shape
this capacity. In our simulations, we derive the pest-related parameters from historical data con-
cerning the ECB, as Appendix B details. For instance, we assume the initial distribution of larvae
follows a normal distribution with an average of two larvae per plant, reecting ECB statistical
data. Table 4 in Appendix B documents the economic impact of the larvae, providing a detailed
reference for the loss estimations related to ECB infestations.
Additional exogenous parameters include corn prices and the average corn yield, signi-
cantly impacting farmers’ input decisions and protability. We assume that the price of corn and
the average yield are normally distributed around their historical averages, specically $6.40 per
bushel for the price and 173 bushels per acre for yields.
12 13
ese averages serve as the basis
for our simulations, reecting the inherent uncertainties associated with corn production. To ac-
commodate the variability in price and yield, we generate a sample of 10,000 observations from
these normal distributions, thereby incorporating the uncertainties related to price and yield into
our analysis.
11
e median value for the SIRAC network is 1.88, the average for high precision is 5.04 (sd= 2.34), and the average
for low precision signal is 0.95 (sd= 0.45). e Tables D.1, D.2, and D.3 in Appendix D oer descriptive statistics for
the calculated signal precision for our dierent simulation models.
12
We calculated the average corn price using the corn price data for the last two decades sourced from Iowa State
Extension: hps://www.extension.iastate.edu/agdm/crops/pdf/a2-11.pdf.
13
20-year average yield calculated based on the corn yield data from USDA - NASS.
14
5 Simulation Results
5.1 Baseline Scenario
Our simulations begin with a baseline scenario, establishing a reference point for comparing the
outcomes from simulations that include the SIRAC network. is baseline scenario is based on
four simplifying assumptions, which we later relax in subsequent sections. First, we assume that
farmers within the network have limited knowledge and experience regarding the optimal timing
for pest treatment. is acknowledges the challenges farmers may face in making informed pest
management decisions. Second, we categorize the precision of scouting technology into two
levels: low and high precision. ird, we assume that farmers in the network receive ten uniform
informative signals from their peers. Additionally, we dene two precision levels for the network
signal: low and high. In this baseline scenario, we treat the precision of the farmer’s scouting and
network signals as following a standard normal distribution. We calculate low and high precision
values as the averages of signal precision below and above the median.
14
is model does not
consider the geographical distance between signal senders and receivers, implying an equal level
of trust in all signals, regardless of their source. Lastly, we assume that the pest carrying capacity
has a normal distribution with an average of 3 per plant
15
and a standard deviation of 0.5. is
approach provides a simplied framework for assessing the network’s impact, which we expand
upon in the subsequent analyses.
Figure 2 illustrates the simulation results using the baseline model. e blue histograms
across each graph display the distribution of expected gains for farmers who utilize scouting to
gather insights on the pest population, compared to a baseline where decisions are made solely
based on prior knowledge. Meanwhile, e orange histograms show the expected gains for farm-
ers who improved their decision-making with scouting and information obtained through their
network, enhancing their pest management strategies.
Each subgraph within Figure 2 corresponds to a distinct scenario regarding the precision of
information derived from scouting and the network. For example, Figure 2a illustrates outcomes
for a scenario where both scouting and network-derived learning signals have low precision.
is scenario reects a context in which the scouting technology is relatively undeveloped, and
the reliability of information from the network is uncertain. Conversely, Figure 2d showcases
the case where the precision from both scouting and networking is high, indicating advanced
scouting technology and reliable network information. ese distinctions show how varying
information precision levels can impact pest management strategies’ eectiveness.
Within our model, farmers can learn through three methods: their own previous experience;
direct observation and action (scouting); and, exchanging information within a network. Adding
each new learning method improves how eectively farmers can manage pests, reducing their
potential losses. A reduction in losses due to beer pest management when farmers use scouting
methods in their elds represents the benet gained from scouting alone, shown by the blue
histogram. Similarly, the further decrease in losses when employing both methods measures the
advantage of combining scouting with networking, depicted by the orange histogram. erefore,
14
We x the precision of the farmer’s initial knowledge at 0.14, which is the average value for the sample of
farms with precision lower than the median from a uniform distribution. e xed value for the previous knowledge
precision does not aect simulation results because it is constant across simulations for scouting and for scouting
and networking.
15
e assumption is based on the statistics provided by Christian H. Krupke (2017). We revise this assumption for
the simulations for SIRAC and expanded network based on the values suggested in Guse et al. (2002) and Tyutyunov
et al. (2008).
15
the dierence between the gains shown in the orange and blue histograms highlights the extra
value provided by the social network. is dierential quanties the network’s contribution
beyond what scouting alone can oer, underlining the signicance of collaborative learning in
enhancing agricultural practices.
16
(a) Scout precision: Low; Network precision: Low (b) Scout precision: Low; Network precision: High
(c) Scout precision: High; Network precision: Low (d) Scout precision: High; Network precision: High
Figure 2: Simulation of Farmer’s Expected Gains by Signal Precision - Baseline Model
Note: Figure 2 shows the distribution of farmer’s expected gain from learning from scouting and from the network.
Farmers have three channels of learning: previous knowledge, scouting, and network. e blue histograms plot
the distribution of farmer’s gain from scouting relative to the reference case of only previous knowledge. e
orange histograms plot the distribution of farmer’s gain from scouting and networking relative to the case of
only previous knowledge. e dierence between the orange and blue histograms captures the gain from the
network. e dashed vertical line represents the median value of each distribution. Each graph plots distributions
for dierent precision levels of the signals from scouting and from the network.
e baseline model serves as a foundation for comparing scenarios where we take the net-
work’s inuence into account. According to our simulations, the average expected gain from
scouting activities, when utilizing low precision technology, stands at approximately $15 per acre.
However, the range of potential reduction in losses due to scouting varies widely, from as lile as
$0 to as much as $100 per acre, depending on the farm’s specic circumstances (as Figures 2a and
2b illustrate). e lower end of this range corresponds to conditions where a farm experiences no
pest infestation, in which case the value of scouting for gathering information is negligible. On
the other hand, farms facing severe pest challenges realize the greatest advantage from scouting
eorts. Enhancing the precision of scouting technology raises the average expected benet from
scouting activities to $60 per acre. In instances of high vulnerability to pest aacks, the benet
can surge to over $200 per acre (Figures 2c and 2d).
e introduction of access to a network of farmers, adding a third channel of learning, no-
tably alters the distribution of expected gains for farmers. Initially focusing on scenarios where
scouting technology has low precision, our simulations reveal that the expected gain from inte-
16
Our simulation of network gains does not need the specication of a cost for scouting, as any such cost would
cancel out in the computation of the farmer’s expected gains from the network, being incorporated in the orange
and blue histograms.
16
grating scouting with network learning varies signicantly. With network signals of low preci-
sion, the expected gain begins at $35 per acre (as seen in Figure 2a), and increases to $70 per acre
when the network signals are of high precision (illustrated in Figure 2b). e maximum potential
gain exceeds $100 per acre with low precision signals and can surpass $200 per acre with high
precision signals. Transitioning to scenarios where scouting technology is highly precise, the
distribution of expected gains slightly shis right (as Figures 2a and 2b depict). is observation
suggests that while a uniform improvement in scouting technology benets all farms, farmers
also having the capability to learn from their peers somewhat limits the marginal gain of such
improvement.
e dierence between the expected gain from utilizing both scouting and network learning
(indicated by a vertical orange dashed line) and the expected gain from solely relying on scout-
ing (marked by a vertical blue dashed line) quanties the economic value derived from learning
within a farmer network, as depicted in Figure 2. is dierential, representing the network’s
value, is notably higher, approximately $50 per acre, when the precision of the scouting signal
is low, yet the precision of the network signal is high, as illustrated in Figure 2b. Conversely, in
situations where the scouting signal’s precision is high and the network signal’s is low, the in-
cremental benet of incorporating the farmer network into pest management strategies is min-
imal, as shown in Figure 2c. However, the learning value derived from the network increases
signicantly at the distribution’s tail, particularly for farmers who are most at risk from pest in-
festations. For these vulnerable farmers, the value aributed to the network can surpass $200 per
acre.
5.2 Simulation of the SIRAC Network and ECB Pest Management
In this section, we extend our simulation to assess the economic value of a network in a more re-
alistic seing for pest management. We use the actual number of traps installed across 121 farms
within the SIRAC network, along with the pairwise distances among these farms, to determine
the precision of scouting and the precision of network signals. Initially, we assign a unique pre-
cision level for the scouting technology to each farm, based on the real number of traps installed.
We calculate the precision for each farm’s scouting as the ratio of the number of traps to the
maximum observed, which is seven.
Next, we account for variations in the precision of signals received by farmers from their
network peers. is adjustment considers both the scouting precision of the sending farm and the
geographical distance to the receiving farm. In line with the baseline model, each farmer receives
10 signals from within the network, but now, each signal varies in precision. We use a weighted
average, where the distance from the sender to the receiver gives the weigh, to determine the
precision of a signal received by a farmer. Specically, a signal from a nearby farm equipped
with high-precision scouting technology will carry more weight than one from a distant farm
with low scouting precision. is approach allows us to investigate how both technological and
spatial factors inuence the value of information exchanged within the network
Our SIRAC simulations focus on pest management strategies targeting the ECB, primarily
because the ECB’s life cycle aligns well with the pest population dynamics outlined in equation 2.
Our model is adaptable and can be extended to other pests by modifying the population growth
function to t specic pest life cycles.
Table 1 presents all the parameters used in our simulations, including their values and dis-
17
tributions. We tailor these parameters—specically the pest death rate, carrying capacity, and
initial population to the ECB based on empirical data and research ndings detailed in the cited
study. Notably, we ajust the average pest carrying capacity between the baseline scenario and the
SIRAC simulation—while we set it to 3 in the baseline to simplify initial assessments, we revise it
to 22 in the SIRAC simulation to more accurately reect the ECB’s ecological reality and poten-
tial for population growth under optimal conditions. We model the growth rate of the ECB pest
population as a function of GDDs, since its population growth is contingent on the accumulation
of degree-days above the ECB’s developmental threshold temperature of 50°F.
Figure 3 displays the expected gains for farmers in the SIRAC simulation, with the blue and
orange distributions indicating the farmer’s gains with scouting (blue) and with scouting plus net-
working (orange) for varying precision levels of scouting and network signals, consistent with the
baseline model. e overall trends between the baseline and SIRAC simulations remain similar,
highlighting the signicant benets of network participation. Particularly, the dierence in ex-
pected gains between the orange (network plus scouting) and blue (scouting only) distributions,
marked by vertical dashed lines, is notably larger in scenarios where the precision of scouting
technology is lower (as comparisons between gures 3b and c illustrates).
(a) Scout precision: Low; Network precision: Low (b) Scout precision: Low; Network precision: High
(c) Scout precision: High; Network precision: Low (d) Scout precision: High; Network precision: High
Figure 3: Simulation of Farmer’s Expected Gains by Signal Precision - SIRAC Network
Note: Figure 3 shows the distribution of farmer’s expected gain from learning from scouting and from the network
for the SIRAC network with an application to management of ECB pest. Farmers have three channels of learning:
previous knowledge, scouting, and network. e blue histograms plot the distribution of farmer’s gain from
scouting relative to the reference case of only previous knowledge. e orange histograms plot the distribution
of farmer’s gain from scouting and networking relative to the case of only previous knowledge. e dierence
between the orange and blue histograms captures the gain from the network. e dashed vertical line represents
the median value of each distribution. Each graph plots distributions for dierent precision levels of the signals
from scouting and from the network.
However, two dierences emerge between the results of the baseline and SIRAC models.
First, the magnitude of gains from network participation is considerably higher in the SIRAC
model. Specically, with low precision scouting technology, expected gains from network par-
ticipation rise to $70 per acre with low precision network signals (Figure 3a) and to $231 per
18
acre with high precision network signals (Figure 3b). With high precision scouting technology,
the benets from network participation decrease but remain signicant, at $20 and $80 for low
and high precision network signals, respectively. e primary reason for this increased network
value in the SIRAC simulation is the higher pest carrying capacity, adjusted from 3 to 23. is
adjustment suggests a greater likelihood of severe pest infestations, potentially leading to more
substantial yield losses. Consequently, as the potential severity of pest threats escalates, the value
of the network as a learning channel increases.
e second notable distinction in the simulation results for the SIRAC network lies in the
widened spread of the distribution of gains from network participation. We aribute this in-
creased variability among farmers’ gains to the signicant heterogeneity in the precision of
scouting and network signals. Farmers equipped with advanced scouting technologies tend to
benet less from the network, as their existing systems already provide them with a high level of
pest management eciency. Conversely, farmers that have less sophisticated scouting technol-
ogy but have access to more precise signals—potentially due to their proximity to experienced
farmers—stand to gain more from network participation. is variation in gains underscores the
impact of spatial and technological factors on the value derived from the network. Understanding
these dynamics oers valuable insights for optimizing network design.
e Distribution of Network Gains
To identify which farms benet most from network participation, we analyze the expected gains
from network involvement across dierent quantiles of the distribution of gains. Table 2 provides
a detailed look at the economic gains farmers can anticipate from being part of the network,
segmented by quantiles. is analysis combines the results of all scenarios illustrated in the four
graphs of Figure 3. Furthermore, Table 2 includes farm characteristics at each percentile of the
distribution of gains. e characteristics examined include the average pairwise distance between
farms within each percentile, the average number of GDDs, the precision levels of scouting and
network signals, and data on corn prices and yields. ese aributes help identify the factors
contributing to the dierential benets observed across the network.
We observe the smallest expected gain from participation in the SIRAC network at the 5
th
quantile, amounting to $85 per acre (Panel A of Table 2). A greater average distance from their
peers within the network, the lowest accumulation of GDDs, and receiving the least precise net-
work signals characterize farms that benet the least from the network. Additionally, these farms
have the most advanced scouting technology available to them and the lowest average corn prices
and yields. Such farms are initially less susceptible to pest infestations and possess a superior ca-
pability to gather and learn from their own scouting data. Moreover, the information they receive
about pest management from their peers through the network tends to be less accurate, further
diminishing the relative value of network participation for these particular farmers.
e highest gain from participation in the SIRAC network reaches $501 per acre at the 95
th
quantile. Farmers in this high quantile are more vulnerable to pest infestations due to a greater
accumulation of GDDs, and they benet from the highest corn prices and yields. Consequently,
these farms have more at risk in the face of ineective pest management. Additionally, lower
precision in their scouting eorts characterizes farmers at the 95
th
quantile. However, they re-
ceive the most precise signals from the network, partly due to their proximity to other members
of the network. erefore, the farmers who benet the most from the network are most at risk
due to external environmental and market conditions, and are also less capable of independently
acquiring optimal pest management knowledge.
19
Table 2: e Farmer’s Expected Gain from Network Participation
Expected Gain Sender-Receiver Signal Precision Corn Price Corn Yield
from Network Distance (miles) GDD Network Scouting ($ per bushel) (bushels per acre)
$ per acre (St.Dev) Avg. P10 Avg. P90 Avg. P90 Avg. P90 Avg. P90 Avg. P90
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
A. SIRAC Network
Q5 85 (17) 119 28.04 1,816 2176 0.21 0.18 0.62 0.86 5.97 7.05 1.43 2.01
Q25 149 (22) 116 25.97 1,847 2226 0.28 0.68 0.62 0.86 6.21 7.26 1.80 2.43
Q50 231 (26) 112 19.74 1,868 2266 0.22 0.68 0.60 0.86 6.30 7.35 1.92 2.53
Q75 325 (30) 110 16.37 1,888 2300 0.48 0.68 0.59 0.71 6.43 7.45 1.62 2.62
Q95 501 (114) 109 14.79 2,004 2460 0.51 0.68 0.58 0.57 6.70 7.71 2.29 2.85
B. Expanded Network
Q5 93 (19) 116.20 37.00 1,802 2155 0.26 0.21 0.76 1.00 5.93 6.97 1.38 1.92
Q25 163 (25) 110.60 32.00 1,847 2230 0.34 1.00 0.76 1.00 6.21 7.24 1.84 2.39
Q50 252 (28) 103.80 29.00 1,865 2260 0.56 1.00 0.76 1.00 6.41 7.48 2.00 2.65
Q75 353 (32) 96.93 29.85 1,891 2322 0.69 1.00 0.75 1.00 6.37 7.42 2.00 2.58
Q95 539 (117) 93.21 29.35 2,007 2459 0.75 1.00 0.76 1.00 6.66 7.69 2.23 2.79
C. Expanded Network with Signal Selection
Q5 148 (30) 7.04 2.23 1,809 2164 0.96 1.00 0.75 1.00 5.89 6.99 1.39 1.98
Q25 247 (31) 6.93 2.23 1,821 2198 0.96 1.00 0.75 1.00 6.15 7.25 1.86 2.29
Q50 347 (30) 6.83 2.23 1,848 2237 0.97 1.00 0.75 1.00 6.39 7.41 1.94 2.69
Q75 461 (38) 6.76 2.23 1,898 2313 0.97 1.00 0.76 1.00 6.32 7.49 2.29 2.62
Q95 688 (109) 6.69 2.20 2,030 2501 0.97 1.00 0.76 1.00 6.65 7.58 2.33 2.76
Panel A summarizes the farmer’s expected gains from participating in the SIRAC network for ECB pest management. Panel B reports gains for the
expanded network. Panel C reports gains for the expanded network with signal selection. e table reports averages, standard deviations and tail vales
at the 10
th
percentile (P10) and the 90
th
percentile (P90).
20
Network Expansion
To evaluate the benets of expanding the SIRAC network, we simulate an enhanced version of it.
In this version, we introduce ve additional neighboring farms within a 30-miles radius for each
of the 121 existing farms in the SIRAC network, resulting in a total of 605 farms. Additionally,
we randomly assign a varying number of traps to each of the newly added farms. e simulation
parameters remain consistent with those used in the previous SIRAC simulation. Figure 1 presents
a map illustrating this expanded network in Iowa.
17
Table 2, Panel B, shows the outcomes of the simulation for the expanded network, focusing
on ve quantiles in the distribution of farmers’ gains from participating in the network. As an-
ticipated, the simulation reveals that the expanded network brings farmers geographically closer
to each other across all quantiles, as indicated by the reduced pairwise distance (Column 2). is
proximity enhances the precision of the information signal within the expanded network (Col-
umn 6). A notable benet of this larger network is the increased accessibility to peer farmers
situated closer by.
e simulation of the expanded network reveals a uniform 9% increase in farmers’ expected
gains across all quantiles, underscoring the positive impact of enhanced signal precision in a
larger network (Table 2 Panel B, Column 2).
18
is improvement in gains reects the enhanced
quality of signals concerning pest infestations, derived from closer neighbors. Such proximity
enables farmers to ne-tune their pest management strategies more eectively. However, the
benet observed from expanding the network to ve times its original size is smaller than ex-
pected. e key issue lies in the random selection of network signals—despite an increase in
the number of peer farmers, some of whom are now closer, the process does not prioritize sig-
nals based on the geographical proximity between sender and receiver. Consequently, farmers
may still receive signals from peers several hundred miles away. Without a rened approach
to selecting signals, the modest gains from network expansion are primarily due to a slightly
higher chance of receiving a more accurate signal. To address this limitation, we next simulate
the expanded network incorporating a mechanism for selective signal reception.
Expanded Network with Signal Selection
In this section, we assess the advantages of network expansion coupled with signal selection. Un-
like the previous setup, farmers now exclusively receive signals from the 10 nearest peer farmers,
ensuring that the information is geographically relevant. All other simulation parameters re-
main consistent with the earlier simulation. Panel C of Table 2 presents the results of this rened
simulation approach. is adjustment aims to enhance the precision and applicability of the in-
formation exchanged within the network, potentially leading to more signicant gains for the
farmers by focusing on the proximity of their connections.
With the introduction of signal selection based on geographic proximity, the average dis-
tance for received signals dramatically decreases by over 90%. Specically, at the 95
th
quantile,
the average distance for a signal in the expanded network, which stood at 93.21 miles without
signal selection, decreases to 6.69 miles when implementing signal selection (as Panel C, Column
2 shows). is signicant reduction in distance leads to a 29% increase in the precision of the
17
We further investigated the distribution’s tail by examining statistics for the 90
th
percentile within each quantile
of the distribution. Table 2 includes these extreme statistics for farm characteristics. Farms at the far right tail of
the distribution are, on average, 14.79 miles away from their peers, have experienced 2,460 cumulative GDDs, and
have received network signals that are over three times more precise than those received by farmers at the lower
quantiles.
18
Figure 5 in Appendix D shows the distribution of gains for the expanded network.
21
network signal.
e impact of signal selection is particularly large at the lower quantiles. For example, at
the 5
th
quantile, signal precision increases from 0.26 without signal selection to 0.96 with signal
selection, marking close to a threefold improvement. is suggests that the lower quantiles ben-
et the most from this signal selection methodology. is approach, which prioritizes proximity
over other farm characteristics, enhances the relevance of the information exchanged within the
network.
e introduction of signal selection signicantly boosts the benets farmers derive from
network participation. Specically, at the lowest quantile (the 5
th
), gains escalate from $93 per
acre to $148 per acre, a 59% increase. For the highest quantile, gains increase from $539 per
acre to $688 per acre, a 28% improvement. A comparison of the gains under the original SIRAC
network against those achieved with the expanded network incorporating signal selection reveals
even more striking enhancements—a 34% increase at the highest quantile and a 74% at the lowest
quantile.
ese ndings highlight the critical importance of signal selection in network design, demon-
strating that even a basic criterion for signal selection can lead to substantial economic benets.
Furthermore, we can rene the method for selecting signals, such as by incorporating additional
farm characteristics, including the accuracy of peer farmers’ scouting technology.
5.3 Extreme Heat Simulation: e Network Adaptation Value
In this section, we explore the farmer’s network’s potential to mitigate yield losses from acceler-
ated pest infestations caused by climate change. We dene the adaptation value of the network
as the additional economic gain from network participants in a scenario where the number of
GDDs increases due to climate change. is adaptation value stems from two key mechanisms.
First, the network functions as an early-warning system for pest infestations triggered by
a warmer climate. Second, the uncertainty regarding the optimal timing of pesticide application
tends to rise with higher degree days, owing to the spatial variability in climate change. Even
within a state, certain areas may be aected dierently during warmer seasons. As the challenge
of managing pests becomes more complex for farmers, the ability to learn from peers becomes
increasingly valuable.
Climate change can accelerate the growth rates of pest populations in a given location and
facilitate the emergence of pests that are more prevalent in warmer climates Bale et al. (2002),
Fand et al. (2012), and Skend
ˇ
zi
´
c et al. (2021). For instance, in the case of the ECB, an increase
in GDDs can result in the early emergence of the rst occurrence of the pest, a swier growth
in pest population, and an overall rise in the number of ECB generations on a farmKocm
´
ankov
´
a
et al. (2010), Gagnon et al. (2019), Gagnon et al. (2019), Skend
ˇ
zi
´
c et al. (2021), and Schneider et al.
(2022).
Researchers from Iowa State Extension have shown that there can be up to four generations
of ECB during a single season in warmer southern states. In the corn belt, however, there are
typically two or three generations of ECB in a season. Failure to manage the rst generation
of ECB in a timely manner not only increases the damage caused by the initial generation but
also raises the risks of further losses from subsequent ECB generations. Managing the ECB pest
22
promptly becomes even more important in warmer climates
19
.
We assess how participation in agricultural networks can serve as an adaptation strategy to
climate change by examining three distinct scenarios that project increases in GDD by 10%, 20%,
and 30%. ese scenarios draw upon historical data observed by the Environmental Protection
Agency (EPA) in the United States, which documents a signicant rise in GDD nationwide from
1984 to 2020
20
. e EPA’s ndings reveal an average increase of 9% in GDD over this 36-year
timeframe, with certain regions experiencing jumps of over 20%. is analysis aims to understand
the adaptive benets that network participation might oer in response to varying degrees of
climate-induced changes in agricultural conditions.
Our simulation specically targets the extreme value of GDDs under each climate change
scenario to evaluate the maximum potential of the network for adapting to and mitigating severe
pest infestations. We characterize extreme GDDs as values exceeding two standard deviations
from the mean. Given the nonlinear increase of pest population growth rates with GDDs, we
predict only moderate adaptation benets from within-network learning at median GDDs values.
We verify this prediction in simulations reecting an average increase in the median number of
GDDs
21
. Furthermore, it is important to note that pest carrying capacity, which is the maximum
pest population that can survive given the environmental and ecological constraints, naturally
limits the impact of GDDs on pest populations. erefore, we anticipate that the adaptive benets
provided by the network participation will likely diminish at higher GDD values. Our simulations
aim to explore these boundaries, identifying the point at which the network’s adaptive benets
start to decrease as GDD increase.
In our climate change simulation, we adopt a distinct approach for measuring the adaptation
value of the network, diverging from the methods used in our initial simulations. To quantify the
network adaptation value, we employ a dierences-in-dierences (DiD) strategy. is process
involves two primary steps:
1. First Dierence: We start by computing the dierence in gains from scouting activities,
with and without the impact of climate change, across 10,000 simulations. is represents the
initial variation in outcomes aributable to climate change alone.
2. Second Dierence: Next, we calculate the gains from combining scouting and networking
activities, both with and without the inuence of climate change. is step assesses the combined
eect of networking and scouting in the context of climate change.
We then determine the adaptation value of the network by the dierence between these
two measures: the gain from combining scouting and networking versus the gain from scouting
alone. Essentially, our outcome variable in the climate change simulations reects the expected
gain from participation in networks, contrasting conditions with and without climate change,
specically focusing on the upper tail of the GDDs distribution.
Figure 4 shows the results of climate change simulations for the expanded network, focusing
19
Ecology and management of ECB in Iowa eld corn, Iowa State Extension, 2017:
hps://store.extension.iastate.edu/product/15141
20
Percentage change in growing degree days 1948-2020. Source: hps://www.epa.gov/climate-
indicators/climate-change-indicators-growing-degree-days. Data source: NOAA, 2021. NOAA (National
Oceanic and Atmospheric Administration). 2021. Global Historical Climatology Network Daily: Data access.
hps://www.ncei.noaa.gov/products/land-based-station/global-historical-climatology-network-daily. Accessed
March 2021.
21
Simulation results spanning the entire distribution of GDDs under the three climate change scenarios are avail-
able upon request from the authors.
23
on a scenario that predicts a 10% increase in GDDs.
22
In this gure, the blue histograms illustrate
the distribution of dierences in farmers’ expected gains from scouting activities alone. is
comparison is made between the scenario with a 10% increase in GDDs and the baseline scenario,
which assumes no change in climate. Conversely, the orange histograms depict the distribution
of dierences in farmers’ expected gains when integrating scouting and network signals, with
the same comparison between the post-10% GDDs increase scenario and the climate unchanged
baseline.
Figure 4 shows the climate change simulation results for the expanded network for the cli-
mate change scenario with an 10% increase in GDDs. e blue histograms represent the distribu-
tion of dierences in farmers’ expected gains from scouting alone, comparing the scenario aer
a 10% increase in GDDs to the baseline scenario without climate change. Meanwhile, the or-
ange histograms show the distribution of dierences in farmers’ expected gains from combining
scouting and network signals, again comparing the post-10% GDDs increase scenario to the no
climate change baseline. e dashed vertical lines in each graph mark the median value of the
distributions.
(a) Scout precision: Low; Network precision: Low (b) Scout precision: Low; Network precision: High
(c) Scout precision: High; Network precision: Low (d) Scout precision: High; Network precision: High
Figure 4: Simulation of the Network Climate Change Adaptation Value - Expanded Network
Note: Figure 4 illustrates the distribution of farmers’ expected gains from both the expanded network and scouting
activities under the scenario of a 10% increase in GDDs, specically focusing on the management of ECB pests. e
blue histograms represent the distribution of dierences in farmers’ expected gains from scouting alone, comparing
the scenario aer a 10% increase in GDDs to the baseline scenario without climate change. Meanwhile, the orange
histograms show the distribution of dierences in farmers’ expected gains from combining scouting and network
signals, again comparing the post-10% GDDs increase scenario to the no climate change baseline. e dierence
between the orange and blue histograms quanties the network’s adaptation value under the scenario of a 10%
GDDs increase. is dierence highlights the additional benet that network participation oers over scouting
alone in adapting to climate change impacts. e dashed vertical lines in each graph mark the median value of the
distributions. Each graph within Figure 4 shows distributions for various precision levels of scouting information
and network signals.
Graph A of Figure 4 shows the adaptation value of the expanded network under conditions
of extreme GDDs, specically when both scouting technology and network signal precision are
22
Appendix D presents the simulation results for the original SIRAC network.
24
low. e key metric for assessing adaptation value is the dierence between the mean values of
the orange and blue distributions, denoted by dashed vertical lines. In this scenario, the expected
adaptation value of the network is approximately $40 per acre, or around 40% of the expected
network gain under normal climatic conditions. is result shows that, even with low precision
in learning mechanisms, the network still oers signicant value in scenarios characterized by
extreme GDDs and a heightened risk of severe pest infestations.
Graph B of Figure 4 illustrates a scenario in which the precision of the network signals
has been enhanced, leading to an increase in the adaptation value of the expanded network to
$60 per acre. is increase in adaptation value highlights the importance of the network for
farmers facing potentially severe pest infestations, especially when other reliable sources of pest
management information are lacking. By comparing the outcomes presented in Graphs A and
B, we can quantify the benets of enhancing network signal precision across all farms. e
dierence, representing an expected gain of approximately $20 per acre, represents the value
derived from investing in the improvement of network signal precision.
Graphs C and D from Figure 4 present the outcomes of simulations where scouting tech-
nology precision is uniformly high across all farms within the network. Although the real-world
likelihood of every farm having access to such high-precision scouting is small, analyzing this
scenario is informative about the lower bound for the network’s adaptation value.
In scenarios where farms have advanced internal capabilities for monitoring pest popula-
tions, the incremental benet of external information received from network peers naturally di-
minishes. e simulation results presented in Graph C, where the expected adaptation value of
the expanded network—given high precision in scouting technology but low precision in network
signals—is relatively modest, at about $25 per acre, reects this phenomenon. When we enhance
the precision of the network signal, the expected adaptation value of the network sees only a
slight increase to approximately $40 per acre, as shown in Graph D. ese ndings highlight that
even under a more conservative scenario where all farms have high scouting technology, there
remains a discernible but marginal adaptation value in learning from network peers.
e Distribution of Network Adaptation Values
Table 3 details the adaptation values associated with network participation across ve quan-
tiles, considering three climate change scenarios (GDD + 10%, GDD + 20%, and GDD + 30%)
and three dierent networks (SIRAC, expanded network, and expanded network with signal se-
lection). e simulation focuses on extreme GDD within each climate change scenario. Table
t:climatechangesim includes the corresponding GDD for each quantile of the adaptation value
distribution.
A signicant nding from this analysis is the substantial variation in adaptation values
across the distribution for each network simulation and climate change scenario. At the lower
end of the spectrum, adaptation values are relatively modest, ranging from $10 to $42 per acre
across the various climate change scenarios for the SIRAC network, as noted in Panel A of Table
t:climatechangesim. In stark contrast, at the highest quantiles, the adaptation value for the sce-
nario with a 10% increase in GDDs climbs to $588 per acre. is value further escalates to $828
per acre under the more severe climate change scenario.
We can primarily aribute the variation in adaptation values across quantiles to two factors:
the initial pest population levels and the magnitude of extreme GDD. Other simulation parame-
25
ters, such as corn prices and yields, remain consistent across quantiles. ese results highlight
the role of farmer networks in providing adaptive benets under scenarios of heightened climate
stress, particularly when the risk of severe pest infestations is elevated.
e simulation results for the expanded SIRAC network show only marginal increases in
adaptation values compared to the original SIRAC network, aligning with our observations under
normal climatic conditions. is outcome, which Panel B of Table t:climatechangesim details,
suggests that merely expanding the number of farms within a network—without addressing the
variability in the precision of information (signals) shared among network members—yields only
modest enhancements in the network’s adaptation value. is nding underscores the limited
eectiveness of network expansion as a standalone strategy for improving adaptation to climate
change.
Table 3: Network Adaptation Value for Extreme Heat Scenarios
GDD + 10% GDD + 20% GDD + 30%
Adaptation Value Extreme Adaptation Value Extreme Adaptation Value Extreme
($ per acre) GDD ($ per acre) GDD ($ per acre) GDD
A. SIRAC Network
Q5 9 (5) 2,775 20 (9) 2,880 24 (13) 3,043
Q25 74 (4) 2,767 124 (6) 2,925 158 (6) 3,067
Q50 154 (7) 2,818 221 (5) 2,944 272 (6) 3,099
Q75 270 (8) 2,914 331 (7) 3,055 378 (8) 3,155
Q95 413 (17) 3,062 451 (8) 3,110 466 (7) 3,197
B. Expanded Network
Q5 9 (4) 2,744 28 (6) 2,878 46 (5) 3,010
Q25 79 (3) 2,784 127 (5) 2,937 173 (8) 3,077
Q50 111 (7) 2,815 210 (9) 2,956 312 (14) 3,094
Q75 264(13) 2,900 421 (14) 3,052 574 (21) 3,199
Q95 561 (39) 3,087 816 (61) 3,183 873 (33) 3,311
C. Expanded Network with Signal Selection
Q5 10 (1) 2,738 25 (4) 2,875 61 (8) 3,016
Q25 81 (0) 2,781 158 (7) 2,918 249 (11) 3,050
Q50 166 (6) 2,812 298 (10) 2,981 439 (10) 3,108
Q75 325 (10) 2,876 486 (13) 3,012 678 (22) 3,171
Q95 680 (9) 3,104 916 (9) 3,215 1128 (69) 3,448
Note: Table 3 presents the simulation results for the adaptation value of network participation by quantiles for
three climate change scenarios. e adaptation value is the additional expected gain of network participation under
a climate change scenario. All simulation results are for the extreme GDD within each climate change scenario.
Extreme GDD is dened as GDD two standard deviations above the mean of the distribution.
However, the introduction of a signal selection mechanism, which prioritizes signals based
on geographical proximity, marks a noticeable improvement in the network’s adaptation capabil-
ities. With this mechanism in place, the adaptation value of the network, especially at the higher
quantiles of the distribution, sees a considerable increase. Notably, under a climate change sce-
nario that projects a 20% increase in GDDs, the adaptation value for the network employing signal
selection jumps to $911 per acre (Panel C of Table t:climatechangesim). is improvement is par-
26
ticularly pronounced in the top quantiles, highlighting the benets of targeted signal selection in
enhancing the network’s adaptation value in the face of more severe climate-induced challenges.
e simulation outcomes for scenarios of extreme heat reveal the ecological constraints that
naturally limit the impact of climate change on pest infestations. ese constraints are primarily
dictated by the pest carrying capacity, which serves as an upper threshold for pest population
growth. Beyond this ecological limit, further increases in temperature do not signicantly exac-
erbate potential losses from pest infestations, nor do they substantially enhance the adaptation
benets of network participation.
is dynamic is evident in the progression of adaptation values across varying degrees of
climate change severity. e adaptation value sees more signicant increases as scenarios shi
from moderate (GDD + 10%) to severe (GDD + 20%). However, the transition from a severe to an
extreme climate change scenario (GDD + 30%) does not yield a proportional increase in adaptation
value. is paern suggests that there is a diminishing return on the adaptation value of network
participation as climate change intensies beyond certain ecological thresholds for pest growth.
6 Conclusions
We assess the economic value of farmer networks in enhancing pest management by adapting
an economic model of learning to pest management and simulating this adapted model across
variations of the SIRAC network. Our ndings reveal considerable variability in the network’s
value, both under typical climate conditions and during extreme heat events caused by climate
change. Networks prove especially benecial for farmers most at risk of pest infestations, with
their value in mitigating the impacts of extreme heat on pest infestations exceeding $900 per acre.
is analysis provides insights for policymakers and businesses aiming to foster the de-
velopment and expansion of such networks. We identify three primary observations from our
simulations that could guide the design of future networks and suggest directions for additional
research:
Variable Network Gains: e gain from network participation varies widely among farm-
ers, indicating the potential for dierentiated pricing strategies. Some farmers might pay more
for network access, while others may require subsidies. Simulations could help determine optimal
pricing strategies.
Strategic Network Expansion and Signal Selection: Expansion benets are limited with-
out signal selection. Our ndings suggest that enhancing the network with a thoughtful signal
selection mechanism, potentially based on geographical proximity and other farm characteristics
(e.g., crop rotation, climate, soil aributes), could maximize the network’s value. Future simu-
lations could explore which farms would benet most from joining the network based on these
rened criteria.
Complementary Role with Insurance: e network’s role as an early warning system
for pest infestations could synergize with agricultural insurance policies, especially as climate
change leads to warmer growing seasons. By mitigating extreme outcomes, the network could
reduce potential losses and, consequently, insurance premiums. Extending the model to include
multiple pests and their distinct life cycles could reveal even greater economic benets of network
participation.
27
Acknowledgments: e authors would like to thank the researchers who contributed to the
SIRAC project at the annual meetings. eir comments and suggestions greatly improved this
paper. Special thanks go to Asheesh Singh for his ongoing support and helpful contributions.
Additionally, the authors are grateful for the assistance provided by supporters from the Iowa
Soybean Association (ISA), including Peter Kyveryga, Mahew Carroll, and Aaron Prestholt, who
provided signicant support for this research. e authors also acknowledge the nancial sup-
port received from the U.S. National Science Foundation (NSF) and Hatch funding from the Center
for Agricultural and Rural Development(CARD) at Iowa State University which helped make this
research possible.
28
References
Adjognon, G. S., Huy, T. N., Gutho, J., and van Soest, D. (2022). Incentivizing social learning for
the diusion of climate-smart agricultural techniques.
Antle, J. M. and Pingali, P. L. (1994). Pesticides, productivity, and farmer health: A philippine case
study. American Journal of Agricultural Economics, 76(3):418–430.
Ariel BenYishay, A. M. M. (2019). Social learning and incentives for experimentation and com-
munication. e Review of Economic Studies, 86(3):976–1009.
Babcock, B. A., Lichtenberg, E., and Zilberman, D. (1992). Impact of damage control and quality
of output: Estimating pest control eectiveness. American Journal of Agricultural Economics,
74(1):163–172.
Bale, J. S., Masters, G. J., Hodkinson, I. D., Awmack, C., Bezemer, T. M., Brown, V. K., Buereld,
J., Buse, A., Coulson, J. C., Farrar, J., et al. (2002). Herbivory in global climate change research:
direct eects of rising temperature on insect herbivores. Global change biology, 8(1):1–16.
Bandiera, O. and Rasul, I. (2006). Social networks and technology adoption in northern mozam-
bique. e economic journal, 116(514):869–902.
Banerjee, A., Chandrasekhar, A. G., Duo, E., and Jackson, M. O. (2013). e diusion of micro-
nance. Science, 341(6144):1236498.
Beaman, Lori, e. a. (2019). Social networks and the dynamics of labor market outcomes: Evidence
from refugees reseled in the u.s. American Journal of Agricultural Economics.
Beaman, L., BenYishay, A., Magruder, J., and Mobarak, A. M. (2021). Can network theory-based
targeting increase technology adoption? American Economic Review, 111(6):1918–43.
BenYishay, A. and Mobarak, A. M. (2019). Social learning and incentives for experimentation and
communication. e Review of Economic Studies, 86(3):976–1009.
Bloomeld, J., Williams, R., Gooddy, D., Cape, J., and Guha, P. (2006). Impacts of climate change
on the fate and behaviour of pesticides in surface and groundwater—a uk perspective. Science
of the total Environment, 369(1-3):163–177.
Carpentier, A. and Weaver, R. D. (1997). Damage control productivity: why econometrics maers.
American journal of agricultural economics, 79(1):47–61.
Carrasco-Tauber, C. and Mo, L. J. (1992). Damage control econometrics: functional specica-
tion and pesticide productivity. American Journal of Agricultural Economics, 74(1):158–162.
Chambers, R. G. and Lichtenberg, E. (1994). Simple econometrics of pesticide productivity.
American Journal of Agricultural Economics, 76(3):407–417.
Christian H. Krupke, Larry W. Bledsoe, J. L. O. (2017). European corn borer in eld corn.
Department of Entomology, Purdue University- Extension.
Conley, T., . U. C. (2001). Social learning through networks: e adoption of new agricultural
technologies in ghana. American Journal of Agricultural Economics, 83(3):668–673.
Conley, T. and Christopher, U. (2001). Social learning through networks: e adoption of new
agricultural technologies in ghana. American Journal of Agricultural Economics, 83(3):668–
673.
Conley, T. G. and Udry, C. R. (2010). Learning about a new technology: Pineapple in ghana.
American Economic Review, 100(1):35–69.
Crane-Droesch, A. (2018). Technology diusion, outcome variability, and social learning: ev-
idence from a eld experiment in kenya. American Journal of Agricultural Economics,
100(3):955–974.
Di Falco, S., Doku, A., and Mahajan, A. (2020). Peer eects and the choice of adaptation strategies.
Agricultural Economics, 51(1):17–30.
29
Fand, B. B., Kamble, A. L., and Kumar, M. (2012). Will climate change pose serious threat to
crop pest management: A critical review. International journal of scientic and Research
publications, 2(11):1–14.
Feder, G. and Savastano, S. (2006). e role of opinion leaders in the diusion of new knowledge:
e case of integrated pest management. World Development, 34(7):1287–1300.
Fernandez-Cornejo, J., Jans, S., and Smith, M. (1998). Issues in the economics of pesticide use in
agriculture: a review of the empirical evidence. Applied Economic Perspectives and Policy,
20(2):462–488.
Foley, J. A., Ramankuy, N., Brauman, K. A., Cassidy, E. S., Gerber, J. S., Johnston, M., Mueller,
N. D., O’Connell, C., Ray, D. K., West, P. C., et al. (2011). Solutions for a cultivated planet.
Nature, 478(7369):337–342.
Foster, A. D., . R. M. R. (1995). Learning by doing and learning from others: Human capital and
technical change in agriculture. Journal of Political Economy, 103(6):1176–1209.
Gagnon, A.-
`
E., Bourgeois, G., Bourdages, L., Grenier, P., and Blondlot, A. (2019). Impact of climate
change on ostrinia nubilalis (lepidoptera: Crambidae) phenology and its implications on pest
management. Agricultural and Forest Entomology, 21(3):253–264.
Grovermann, C., Schreinemachers, P., and Berger, T. (2013). antifying pesticide overuse from
farmer and societal points of view: An application to thailand. Crop Protection, 53:161–168.
Guse, C., Onstad, D., Buschman, L., Porter, P., Higgins, R., Sloderbeck, P., Cronholm, G., and
Peairs, F. (2002). Modeling the development of resistance by stalk-boring lepidoptera (cram-
bidae) in areas with irrigated transgenic corn. Environmental entomology, 31(4):676–685.
Gutierrez, A. P., Ponti, L., d’Oultremont, T., and Ellis, C. (2008). Climate change eects on poik-
ilotherm tritrophic interactions. Climatic change, 87(Suppl 1):167–192.
Hall, G. V., D’Souza, R. M., and Kirk, M. D. (2002). Foodborne disease in the new millennium: out
of the frying pan and into the re? Medical Journal of Australia, 177(11):614–618.
Hodgson, E. and Rice, M. (2017). Ecology and management of european corn borer in iowa eld
corn. ames (ia): Iowa state university extension.[cited 2019 october 18].
Hurley, T. M. and Sun, H. (2019). Soening shock and awe pest management in corn and soybean
production with ipm principles. Journal of Integrated Pest Management, 10(1):7.
Jackson, L. E., Wheeler, S., Hollander, A., O’Geen, A., Orlove, B., Six, J., Sumner, D. A., Santos-
Martin, F., Kramer, J., Horwath, W. R., et al. (2011). Case study on potential agricultural re-
sponses to climate change in a california landscape. Climatic Change, 109:407–427.
Jovanovic, B. and Nyarko, Y. (1995). A bayesian learning model ed to a variety of empirical
learning curves. Brookings Papers on Economic Activity. Microeconomics, 1995:247–305.
Kocm
´
ankov
´
a, E., Trnka, M., Eitzinger, J., Formayer, H., Dubrovsk
`
y, M., Semer
´
adov
´
a, D.,
ˇ
Zalud, Z.,
Juroch, J., and Mo
ˇ
zn
`
y, M. (2010). Estimating the impact of climate change on the occurrence
of selected pests in the central european region. Climate Research, 44(1):95–105.
Krishnan, P. and Patnam, M. (2014). Neighbors and extension agents in ethiopia: Who maers
more for technology adoption? American Journal of Agricultural Economics, 96(1):308–327.
Lansink, A. O. and Carpentier, A. (2001). Damage control productivity: An input damage abate-
ment approach. Journal of Agricultural Economics, 52(3):11–22.
Lazarus, W. F. and Dixon, B. L. (1984). Agricultural pests as common property: Control of the
corn rootworm. American Journal of Agricultural Economics, 66(4):456–465.
Lence, S. H. and Singerman, A. (2023). When does voluntary coordination work? evidence from
area-wide pest management. American Journal of Agricultural Economics, 105(1):243–264.
30
Macdonald, R. W., Harner, T., and Fyfe, J. (2005). Recent climate change in the arctic and its
impact on contaminant pathways and interpretation of temporal trend data. Science of the
total environment, 342(1-3):5–86.
Maertens, A. and Barre, C. B. (2013). Measuring social networks’ eects on agricultural tech-
nology adoption. American Journal of Agricultural Economics, 95(2):353–359.
Miraglia, M., Marvin, H., Kleter, G., Bailani, P., Brera, C., Coni, E., Cubadda, F., Croci, L., De San-
tis, B., Dekkers, S., et al. (2009). Climate change and food safety: an emerging issue with special
focus on europe. Food and chemical toxicology, 47(5):1009–1021.
Miranowski, J. A. (2016). Intervention to manage pest resistance: community-based or govern-
ment regulation. Choices, 31(4):1–8.
M
¨
uller, R., Bandow, C., Seeland, A., Fennel, D., Coors, A., Ebke, P. K., F
¨
orster, B., Martinez-Arbizu,
P., Moser, T., Oetken, M., et al. (2010). Bik f adamus: a novel research project studying the
response and adaptive potential of single species and communities to climate change in com-
bination with other stressors. Journal of Soils and Sediments, 10:718–721.
Munshi, K. (2004). Social learning in a heterogeneous population: technology diusion in the
indian green revolution. Journal of development Economics, 73(1):185–213.
NASS (2014). 2021 agricultural chemical use survey. United States Department of Agriculture,
National Agricultural Statistics Service, 2022-1.
Noyes, P. D., McElwee, M. K., Miller, H. D., Clark, B. W., Van Tiem, L. A., Walco, K. C., Erwin,
K. N., and Levin, E. D. (2009). e toxicology of climate change: environmental contaminants
in a warming world. Environment international, 35(6):971–986.
Rebaudo, F. and Dangles, O. (2011). Coupled information diusion–pest dynamics models predict
delayed benets of farmer cooperation in pest management programs. PLoS Computational
Biology, 7(10):e1002222.
Ronald B. Hammond, A. M. and Eisley, J. B. (2014). Corn rootworm management. Ohio State
University Extension Bulletin.
Roos, J., Hopkins, R., Kvarnheden, A., and Dixelius, C. (2011). e impact of global warming on
plant diseases and insect vectors in sweden. European Journal of Plant Pathology, 129:9–19.
Rosenzweig, C., Iglesius, A., Yang, X.-B., Epstein, P. R., and Chivian, E. (2001). Climate change
and extreme weather events-implications for food production, plant diseases, and pests.
Schneider, L., Rebetez, M., and Rasmann, S. (2022). e eect of climate change on invasive crop
pests across biomes. Current Opinion in Insect Science, 50:100895.
Singerman, A., Lence, S. H., and Useche, P. (2017). Is area-wide pest management useful? the
case of citrus greening. Applied Economic Perspectives and Policy, 39(4):609–634.
Skend
ˇ
zi
´
c, S., Zovko, M.,
ˇ
Zivkovi
´
c, I. P., Le
ˇ
si
´
c, V., and Lemi
´
c, D. (2021). e impact of climate
change on agricultural insect pests. Insects, 12(5):440.
Takahashi, K., Mano, Y., and Otsuka, K. (2019). Learning from experts and peer farmers about
rice production: Experimental evidence from cote d’ivoire. World Development, 122:157–169.
Tyutyunov, Y., Zhadanovskaya, E., Bourguet, D., and Arditi, R. (2008). Landscape refuges delay
resistance of the european corn borer to bt-maize: a demo-genetic dynamic model. eoretical
Population Biology, 74(1):138–146.
Udry, C. (2010). e economics of agriculture in africa: Notes toward a research program. African
Journal of Agricultural and Resource Economics, 5(311-2016-5540):284–299.
Vasilaky, K. N. and Leonard, K. L. (2018). As good as the networks they keep? improving outcomes
through weak ties in rural uganda. Economic Development and Cultural Change, 66(4):755–
792.
31
Vreysen, M., Robinson, A., Hendrichs, J., and Kenmore, P. (2007). Area-wide integrated pest
management (aw-ipm): principles, practice and prospects. Area-wide control of insect pests:
from research to eld implementation, pages 3–33.
Wang, Y., M
¨
ohring, N., and Finger, R. (2023). When my neighbors maer: Spillover eects in the
adoption of large-scale pesticide-free wheat production. Agricultural Economics, 54(2):256–
273.
Zhengfei, G., Oude Lansink, A., van Iersum, M., and Wossink, A. (2006). Integrating agronomic
principles into production function specication: a dichotomy of growth inputs and facilitating
inputs. American Journal of Agricultural Economics, 88(1):203–214.
32
Appendix A: Learning Model Derivations
Our learning model is constructed based on the target input model, which is a widely used frame-
work in the economics of learning through networks. e model’s core idea is farmers’ oppor-
tunities to learn from their peers about the optimal timing for pesticide application. e optimal
timing for pesticide application is a crucial aspect that farmers seek to learn about, as it directly
impacts their ability to mitigate pest infestations. Our model assumes that this optimal timing
comprises two components.
e rst component, denoted as µ
it
, represents random noise that is unpredictable by farm-
ers. In our model, we assume this noise term follows an independent and identically distributed
(i.i.d.) normal distribution N(0, ϑ
2
µ
). e optimal timing, denoted as τ, represents the most prof-
itable choice for farmers. However, since this is initially unknown, farmers strive to learn and
subsequently select the optimal timing, denoted as τ
∗
, to minimize losses caused by agricultural
pests.
e second component, τ
∗
, represents the universal optimal timing across farms. Given their
individual knowledge and experience levels and prior information, farmers decide the timing of
pesticide application, denoted as τ
it
, and the new information update from learning channels
reduces the noise around optimal timing.
˜τ
i
= τ
∗
+ µ
it
(1)
Farmers update their prior beliefs about the target τ
∗
based on their observations of pest
populations. is process involves using their own observations and information from their peers
to rene their understanding of the optimal timing for pesticide application.
Where µ
i,t
∼ N(0, ϑ
2
µ
). e farmer’s production increases as they learn and choose a treat-
ment time closer to the optimal time τ
∗
.
Let’s denote the population of the pest at the very rst season as P
0
∼ N(µ
P
0
, ϑ
2
P
0
) which
is normally distributed. We also dene the pest population function as:
P
t
= e
r∆GDD
t,t−1
P
t−1
= e
rGDD
P
0
(2)
erefore, P
t
is normally distributed with mean of µe
rGDD
and variance of ϑ
2
P
0
e
2rGDD
. e vari-
able r in the production function indicates the pest’s intrinsic growth rate, which is dened as
the dierence between pest growth g and the pest death rate δ. In a given week the farmers scout
and revise their information on the pest population to determine the treatment timing. For the
observational data we have P
t
∼ N(µ
P
t
, ϑ
2
P
t
) and for the prior we have P
t
∼ N(µ
P
t−1
, ϑ
2
P
t−1
).
We dene the farmer’s production function is dened as follows:
q
i
(τ
i
) = ¯q
i
− αg(GDD)(τ
i
− ˜τ
i
)
2
(3)
is production function measures the deviations from the maximum potential yield stem-
ming from the gap between the optimal time of pesticide application and the farmer’s actual
choice. e second term in the equation can be interpreted as a loss function, which is propor-
tional to the pest growth, and the pest growth is a function of the accumulated growing degree
days. e term αg(GDD) is a loss multiplier proportional to the pest growth rate. e pest
growth rate g(GDD) is a solution for the pest population function dened as dierential equa-
tion
dP
t
dt
= rP
t
(1 −
P
t
κ
). In the pest population function, κ indicates the pest’s carrying capacity.
To derive the optimal timing of the pesticide application, we take the expectation of equa-
tions 1 and 3:
τ
i,w
= E( ˜τ
i
|P
w
) = τ
∗
i
(3)
and,
E(q
i
(τ
i
) = ¯q
i
− αg(GDD)(ϑ
2
˜τ
i
+ ϑ
2
µ
) (4)
33
e equation (4) shows the farmer’s expected gain in production as a function of two vari-
ances. e rst one is the variance of the latest time for pesticide application ϑ
2
˜τ
i
, which farmers
can change by learning and maximizing their prot through learning from their own experience
and also receiving signals from the information neighbors. e second is the noise variance,
which is completely random, and the farmers have no tools to learn and control it. Using Bayes
rule and the farmer’s prior about the latest time of pesticide application, we derive the equation
for the variance. e variance of observational information is ˜τ
i|P
t
∼ N(τ
⋆
, ϑ
2
µ
(1 − ρ
τ,p
)), and
the variance of the prior can be wrien as ˜τ
i,
∼ N(τ
∗
i|P
t−1
, ϑ
2
˜τ
i
|P
t−1
), where ρ
τ,p
is the correlation
coecient between optimal timing ˜τ and the pest population P
t
.
We use the Bayes rule for the variances the derive the posterior variance:
ϑ
2
P
t
=
ϑ
2
P
t
ϑ
2
P
t−1
ϑ
2
P
t
+ ϑ
2
P
t−1
=
e
2rGDD
e
2r
+ 1
ϑ
2
P
0
(5)
erefore, the conditional variance of ˜τ
i
can be calculated as
ϑ
2
˜τ
i
= [1 −
e
2rGDD
e
2r
+ 1
ρ
2
τ |P
0
] ϑ
2
µ
(6)
Using Bayes rule for the variances, we derive the posterior variance ϑ
2
˜τ
i
as follows:
ϑ
2
˜τ
i
=
ϑ
2
˜τ
i|P
t
ϑ
2
˜τ
i|P
t−1
ϑ
2
˜τ
i|P
t
+ ϑ
2
˜τ
i|P
t−1
=
1
1
ϑ
2
˜τ
i|P
t
+
1
ϑ
2
˜τ
i|P
t−1
(7)
If we have N
s
prior observations on pest population from scouting we can rewrite the equation
7 the following equation:
ϑ
2
˜τ
i
=
1
1
ϑ
2
t
0
|P
0
+
N
s
[1−
e
2rGDD
e
2r
+1
ρ
2
τ |P
0
]
1
ϑ
µ
(8)
Dene the precision as : ρ
0
=
1
ϑ
2
t
0
|P
0
and ρ
S
=
1
ϑ
2
µ
we have:
ϑ
2
˜τ
i
=
1
ρ
0
+ γ × ρ
S
(9)
Where γ =
N
t−1
[1−
e
2rGDD
e
2r
+1
ρ
2
τ |P
0
]
.
Equation (10) shows that the farmers have two main channels to reduce the variance. e
rst channel is ρ
0
, the precision of the farmer’s initial estimate of the pesticide application time,
which can be higher for the farmers with more experience and knowledge. erefore, the farmers
with higher experience and knowledge can have a beer initial estimate of the most protable
time of pesticide application, which can translate to higher prot.
e second channel is a channel of learning. ρ
S
is the precision of the farmer’s technology
to learn about the most protable time.
erefore the farmers’ expected gain in production can be wrien as follows:
E(q
i
(τ
i
) = ¯q
i
− αg(GDD)(
1
ρ
0
+ γ × ρ
S
+ ϑ
2
µ
) (10)
Learning through network:
In this section, We derive the farmers’ expected gain from learning through three channels.
We assume that the farmers are connected to a network and have an additional medium to gain
information about the most protable time for pesticide application. Farmer i receives multiple
34
signals from several information neighbors. For simplicity, We assume all signals are similar in
terms of precision level. In the application section we relax this assumption and assume that the
signals are heterogeneous. We also assume that the received signals are not correlated meaning
that cov(µ
i
, µ
j
) = 0, ∀i, j, and cov(µ
j
, µ
−i
) = 0) and farmers make inference about the treatment
time of the pesticide application by updating the priors about the most protable application time.
Following similar steps in the previous section we can derive the variance of the application time
as follows:
ϑ
2
˜τ
i
=
1
ρ
0
+ γ × ρ
S
+ N × ρ
N
(12)
Where N in the equation 12 indicate the number of signals each farmer receives and, ρ
N
shows the precision of the signals and thus, N × ρ
N
, in the denominator reects captures the
learning from network channel.
erefore the value of learning can be measured using the equation 13:
∆ = E(π
∗
i
|with social learning) − E(π
∗
i
|without social learning)
= αg(GDD) [
1
ρ
0
+ γ × ρ
S
+ N × ρ
N
−
1
ρ
0
+ γ × ρ
S
]
(13)
Farmer’s expected gain under climate change:
In the simulations for the climate change scenarios, we create a DiD variable, as explained
in section 5.3. Our DiD outcome variable is dened by simulating equation 13 for the extreme
heat values with and without climate change:
∆
DiD
= ∆
CC
− ∆ (14)
Where ∆
CC
captures the simulated value of the learning from the network for extreme heat
values and ∆ is the simulated value of the learning from the network without climate change.
35
Appendix B - Corn Pests
Corn Rootworm
e western and northern corn rootworms pose signicant challenges to maize production in the
Midwest, causing estimated annual losses exceeding $1 billion in yield and management costs
in the United States. Recent assessments suggest that current losses may be even higher. Corn
rootworm larvae, hatching from late May through early June and feeding until late July, primarily
damage maize by feeding on roots, leading to yield reductions of approximately 15-17% per node
of root damage. is larval feeding can result in severe lodging, further decreasing grain yield
by 11-34%. Research indicates a broad range of yield losses, from 6% to 30%, for one to two
nodes of root injury, with a notable study nding a 17.9% yield loss per node of roots injured.
Under moderate infestation levels, a 15% yield loss could mean a reduction of 30 bushels per
acre, assuming a potential yield of 200 bushels per acre. Management practices, such as selecting
resistant seed varieties, are crucial in controlling corn rootworm populations.
Corn Earworm
e corn earworm, a migratory pest from the southern and southeastern U.S., poses a threat to
Iowa’s corn, especially the second ight arriving in late July. is pest can infest over 50% of
plants in late-planted elds, leading to signicant yield losses. Despite severe injuries reported
in elds planted with pyramided Bt hybrids, resistance in the Midwest is not yet conrmed. In-
secticide applications, guided by pheromone trap catches, are recommended for sweet corn or
late-maturing elds to mitigate losses. e corn earworm’s life cycle is about 30 days, with the
number of generations varying by latitude, indicating a need for vigilant monitoring and timely
interventions.
Western Bean Cutworm
e western bean cutworm, traditionally considered a secondary pest, can cause considerable
economic damage in favorable conditions, with yield losses estimated between 4 and 15 bushels
per acre per larva. Infestations can reduce grain yield by 30 to 40 percent in cases of several
larvae per ear. Scouting for eggs and larvae is critical as bio-tech traits oer limited control, and
the timing of insecticide applications is crucial to prevent larvae from entering the ear and causing
damage. Pheromone trapping is a valuable tool for monitoring adult ights and determining the
optimal timing for insecticide application. Additionally, this pest increases the risk of ear rot
and reduced grain quality, highlighting the importance of monitoring and treating infestations
to maintain grain quality and mitigate economic losses.
European Corn Borer
e European corn Borer (ECB), Ostrinia nubilalis, is an insect primarily found in western Asia
and Europe. e ECB caterpillars feed on the corn plant causing severe yield losses. In the early
1900s, U.S. farmers in Massachuses rst encountered infestations of ECB, which quickly spread
to the Midwest and western states. Iowa farmers rst observed ECB infestations in 1942, leading
to substantial yield losses in the state. Before the introduction of transgenic corn hybrids in
the mid-1990s, ECB infestations resulted in yield losses and pest control costs amounting to one
billion dollars annually for U.S. farmers Hodgson and Rice (2017). e widespread adoption of Bt-
corn, a transgenic corn hybrid, led to a drastic decline in ECB populations throughout the United
States. However, U.S. farmers have been gradually increasing planting of non-Bt corn varieties
to approximately 17% of total U.S. corn production. Furthermore, ECB has become increasingly
more resistant to Bt toxins in corn and to insecticides, particularly in the Midwest were Bt Corn
was widely adopted by farmers (reference). As a result, some areas in the Midwest have witnessed
a resurgence in the ECB population (reference).
36
ECB Population Dynamics and Yield Losses
e lifecycle of ECB consists of four distinct stages: egg, larva (borer or caterpillar), pupa, and
adult (moth). e completion of these stages represents one generation. e larvae of ECB go
through ve molts, or instars, during their development. With each molt, the larvae shed their
skin and increase in size. In the United States, the European corn Borer (ECB) exhibits varying
numbers of generations per season depending on the region and climate. While the ECB can have
up to six generations per season in some areas, the number of generations is typically lower in the
United States, with up to four generations observed. In Iowa, where the climate is inuenced by
summer temperatures, there are typically two generations of ECB each season. However, in years
with longer and warmer summers, there is a possibility of a partial third generation occurring.
e ECB development and the resulting population growth depends signicantly on the
climate, more specically on the accumulation of degree-days above the ECB development tem-
perature of 50ºF. e eggs of ECB are typically laid in irregular clusters containing around 15 to
20 eggs. e developmental threshold for egg hatching is approximately 15°C, and eggs usually
hatch within four to nine days. During the winter, ECB larvae enter a diapause state and sur-
vive by remaining in cornstalks, corn cobs, corn residue, or weed stems. Development resumes
when temperatures rise above 50ºF. Farmers can predict the development of ECB through its four
life stages by tracking the cumulative number of degree days starting from the capture of adults
in the spring using traps. For example, the rst generation of ECB larvae start corn stalk bor-
ing approximately 435 accumulated degree-days aer the detection of the rst spring ECB adult
Hodgson and Rice (2017). e rst instar of the second generation ECB larva occurs with about
1,404 accumulated degree-days.
e ECB larval feeding injuries the corn plant reducing grain production. Table 4 shows
the potential corn yield loss from ECB infestation at dierent stages of corn development. Yield
losses are expressed in terms of percentage reduction in bushels per acreage. e magnitude of
the yield loss is depends on the developmental stage of the corn plant and the average larval
density during an ECB infestation. For example, at the early whorl state, the potential yield loss
from an ECB infestation ranges from 5.5 to 10% as the density of ECB larva increases from one
to three larva per plant (reference here). ECB infestations can potentially reduce corn yields by
12%.
Table 4: Corn Yield Loss from ECB Infestation by Corn Growth Stage
Corn growth stage Number of larva per plant
One larva/plant Two larva/plant ree larva/plant
Early whorl 5.5% 8.2% 10.0%
Late whorl 4.4% 6.6% 8.1%
Pre-tassel 6.6% 9.9% 12.1%
Pollen Shedding 4.4% 6.6% 8.1%
Blister 3.0% 4.5% 5.5%
Dough 2.0% 3.0% 3.7%
Note: Table 4 shows potential corn yield loss computed by Christian H. Krupke 2017 accounting for
physiological stresses but excluding factors such as stalk or ear breakage. Yield losses are expressed
as a percentage of corn yield.
37
ECB Pest Management
Monitoring and Scouting. Farmers can use monitoring methods to predict the ECB population
in their elds. ECB moths can be monitored using black light and pheromone traps, with both
methods providing correlated catches. Pheromone traps specically aract male moths, while
black light traps capture both male and female moths. Although black light traps are generally
more reliable, they can also capture a large number of other insects, requiring extensive sorting
eorts. Among the monitoring techniques, pheromone-baited water pan traps have been found
to be the most ecient in capturing adult moths (reference?). In addition to adult capture meth-
ods, there are alternative techniques to estimate borer phenology. For example, plant phenology,
which refers to the timing of specic developmental stages in plants, can be utilized to predict
corn borer development. ermal summations, which involve calculating accumulated heat units,
have also proven to be highly predictive.
Scouting complements monitoring in helping farmers determine the need for and the timing
of pesticide applications. Trap catches serve as an indicator to initiate thorough scouting in the
eld for the presence of egg masses, as there is only a weak correlation between moth catches and
population density. Hodgson and Rice (2017) highlights the value of scouting for ECP manage-
ment: ”Aempting to manage European corn borer without scouting is oen economically ineective
and may result in wasted application costs.”. However, scouting can also be costly. For rst gen-
eration ECB, farmers must scout for larvae. e Iowa Station Extension recommends ”sampling
from ve representative sets of 20 consecutive plants every 40 to 50 acres”. For second generation
ECB, farmers scout for egg masses to assess the density of the egg population. Farmers must
sample several locations within their elds to estimate the density of eggs and calculate the cost
and benets of pesticides application
23
.
Insecticides. Liquid formulations of insecticides are commonly utilized to protect corn
crops from damage, particularly during the period from early tassel formation until the corn silks
have dried. e recommended application strategies for liquid insecticides can vary, ranging from
a single application prior to silking to weekly applications. Granular formulations have gained
popularity as an alternative to liquid insecticides. ese granules can be placed directly into the
whorl, leading to eective control of rst-generation larvae, as they tend to congregate in this
area. Moreover, insecticide applied in a granular formulation tends to have greater persistence
(reference).
e timing of insecticide application is critical for ECB control. e timing of liquid appli-
cations is typically aligned with egg hatch to prevent infestation. However, if corn borers are
already present in a eld, the critical treatment time shis to just before tassels emerge or at the
moment of tassel emergence from the whorl. is specic growth stage is crucial because the
larvae are actively moving and are more likely to come into contact with the insecticide. For
second generation ECB, the timing of application is also critical as insecticides must be applied
before the larvae enter the corn stalk or ear (reference).
Cultural practices. e destruction of stalks, which serve as the overwintering site for
corn borer larvae, has long been recognized as a critical component of corn borer management
strategies (reference). Merely disking the eld is insucient; plowing to a depth of 20 cm is
necessary to eectively eliminate the larvae. Mowing the stalks close to the soil surface has
proven highly eective, eliminating over 75% of the larvae. Combining mowing with plowing
further enhances the ecacy of larval destruction. Tillage practices must also be considered in
ECB management plans as leaving a signicant amount of crop residue on the soil surface can
actually promote borer survival.
Host plant resistance. Signicant eorts have been devoted to breeding research aimed
at developing resistance against corn borers, particularly in grain corn varieties that face ECB
populations with a single annual generation. One of the key factors contributing to resistance
in seedlings against young corn borer larvae is a chemical compound called DIMBOA. is com-
pound acts as a repellent and deterrent to feeding by the larvae. However, incorporating the
23
Extension agencies such as the Iowa State University Extension department oer farmers templates and exam-
ples of cost-benet analysis for pesticide applications. Such analysis depend not only on the egg density in the elds
but also on an estimate for the expected ECB survivorship in the farm, the price of corn, and the cost of pesticide
applications.
38
known resistance factors into sweet corn varieties has proven to be challenging. For example, in
the case of sweet corn, it is dicult to maintain the desirable taste and texture while enhancing
the plant resistance to ECB (reference).
Bt-corn and resistance management. Bt corn is a genetically modied corn plant engi-
neered to contain genetic material from a toxin produced by the bacterium Bacillus thuringiensis
var. kurstaki. is genetic modication leads to the expression of the toxin, making the plant
toxic to ECB and closely related insects while posing no harm to other animals. e widespread
cultivation of Bt corn has had a profound impact on the population dynamics of ECB (Burkness
et al. (2001) and Hutchinson et al. (2010)). However, overtime the ECB population within regions
with widespread adoption of Bt-corn tend to develop resistance to the Bt toxins. In order to delay
the development of Bt resistance, the industry requires Iowan farmers to create a refugee area
withing their farm with non-Bt corn to induce mating between Bt-suscetible and Bt-resistant in-
sects to delay the evolution process of Bt-resistant insects. Furthermore, farmers must account for
multiple groups of Bt-corn components oering resistance to dierent insects and farmers when
conguring their elds. e adoption of Bt-corn eliminates the need for insecticide applications
reducing management costs but farmers must pay an additional technology fee for Bt-corn seeds.
39
Appendix C - Simulation
Simulations step-by-step
is document outlines the procedures for data preparation and simulation processes. Section C.1
provides detailed information on the data preparation. In Section C.2, the steps for simulating
the value of learning within a small network comprising 121 farmers are explained. Section C.3
elaborates on the steps for the expansion of the network to 605 farmers and how simulations
were conducted using the expanded network. Section C.4 provides details on the simulation of
learning value derived from farmers’ expanded networks with signal selection criteria. Lastly,
Section C.5 describes the simulation steps to simulate learning value within both small and large
networks under various climate change scenarios.
C.1 Preparing the raw data sourced from Iowa Soybeans Association (ISA)
We obtained our raw data from ISA, which is formaed as .csv les. is data contains
details about the number of traps installed on each farm linked to the SIRAC network, with
a total of 121 farms interconnected within the network. In our simulations, we consider
this network as the base network. Additionally, the dataset includes pairwise distances
between farms, measured in meters. e pairwise distance in our small network varies
from 0 to 274 miles with an average of approximately 114 miles and a standard deviation
of 64 miles. We use these observations to create weights for the signal precision. In our
simulations, we select 10 signal senders randomly without restricting the distance from the
signal-receiving farmer. Later, for the signal selection models we will introduce specic
criteria for the signal senders to provide insights on network design. For this section, the
signal senders can be located anywhere within the network. We follow the following steps
to prepare the data for our simulations:
C.1.1 Creating a unique identier for each farm: is step is done in Excel. We as-
sign unique identiers ranging from 1001 to 1121. is is a required step before the
simulations. We use unique identiers to select the signal senders in each round of
simulation. We use the record id in the raw data to assign the unique identiers.
e data also includes variable record id 2 for the paired farm. Each farm is paired
with 120 farms and in total the raw data has 14, 520 observations.
C.1.2 Simulating latitude and longitude: is step is necessary for creating maps to
show the network. For this simulation, we consider the center of Iowa as a reference
farm and simulate the location of other farms based on pairwise distance. To replicate
the simulation, use our Python code to create latitude and longitude for each farm.
is program simulates the latitude and longitude of 121 farms.
C.1.3 Calculate the precision of scouting for each farm: We use the number of traps
installed in each farm to calculate this parameter. Number of traps ranges from 1 to
7. For each farm, we divide the number of traps by the maximum number of traps.
(Example – If there are 4 traps installed in farm A then: Precision of scouting in Farm
A = Number of traps in farm A/7).
C.1.4 Creating weights for signals: We use the pairwise distance for this purpose. First,
we convert the distances from meters to miles. en we assign a weight of 1 if the
pairwise distance is less than or equal to 10 miles. If the pairwise distance is between
10 to 25 miles, we assign a weight of 0.75. If the pairwise distance is 25 to 50 miles,
we assign a weight of 0.50, and nally, if the pairwise distance is more than 50 miles
we assign a weight of 0.25.
C.1.5 Precision of signals: To calculate the precision of the signal we generate an inter-
action variable by multiplying the precision of scouting generated in C.1.3 and the
weights generated in C.1.4 above. Later, in simulations we will use this variable to
calculate the weighted average for the signals received by a farmer.
e output of this step is data with the following variables:
40
Table C.1 - 1: Description of the Variables
Variable Name Description
id sender Unique identier for signal-sending farms
id receiver Unique identier for signal-receiving farms
traps per field sender Number of traps installed in the signal-sending farms
traps per field receiver Number of traps installed in the signal-receiving farms
precision scout Precision of scouting in signal-receiving farms
precision signal Precision of scouting in signal-sending farms
dist miles Distance from signal origins to destination farm (miles)
weight signal Signal weights
w pr sigl Interaction variable (Signal × weights)
Simulated Geographic Coordinates
(Latitude sender , Longitude sender) Simulated geographical location of signal senders
(Latitude receiver , Longitude receiver) Simulated geographical location of signal receivers
Note: e output data should be saved as .csv. e simulations in Python code use the .csv as input data. To
replicate, the Python code for importing the .csv le should be adjusted.
41
C.2 Simulating the value of learning within a small network
We use the data prepared in the previous step C.1, to simulate the value of learning within
the small network. For our simulations, we have developed a Python program. Refer to our
Python code to replicate the results:
C.2.1 Number of iteration: We set the number of simulations to 10001. en, we create a
loop to iterate through the following steps to simulate the value of the network.
C.2.1.1 Assigning values for the exogenous parameters: In this part of the simu-
lation, we dene our exogenous parameters. ese parameters include the dis-
tribution of the pest’s initial population, distribution of pest carrying capacity,
number of scouting, number of signals, distribution of corn price, distribution of
corn yield, and distribution of growing degree days. We use the values outlined
in Table 1 for this purpose. For parameters with distributions, we rst generate a
sample of 10,000 from the distribution, and then randomly select an observation
from the sample.
C.2.1.2 Selecting Signal senders: In each simulation round, we randomly choose 10
farmers to act as signal senders, identied by their unique identiers id
s
ender.
ese selected signal senders are then excluded from the list of potential signal
receivers, identied by their unique identiers id.
C.2.1.3 High vs. Low Precision Signal: In each simulation, every farmer receives
10 signals, each weighted dierently based on its distance from the signal ori-
gin. Table 1 provides details on the weights assigned to each signal. Using these
weights, we compute the weighted average of these 10 signals for each of the 111
signal-receiving farmers.
High precision signal refers to the average of signals with precision greater than
the median value for all of the 111 farmers. Conversely, low precision signal
refers to the average of signals with precision less than the median value. Note:
For the presentation of the results with graphs we used slightly dierent numbers
of simulations for each sub-sample to keep the scales of the graph the same.
C.2.1.4 High vs. Low Precision Scouting: e sub-sample of signal-receiving farmers
with low precision scouting is dened as the average scouting precision for the
farms with the number of traps less than or equal to the median number of traps.
On the other hand, our sub-sample for high-precision scouting is dened based
on the values for scouting precision greater than the median number of pest traps.
C.2.1.5 High vs. Low Initial precision: For low initial precision, we assign a value of
0.14, while for high initial precision, we assign a value of 0.91. In the simulation
of the theoretical model, we determined the 10
th
percentile and 90
th
percentile
of the sample of 10,000 observations from a uniform distribution as the bench-
marks for low and high initial precision, respectively. ese values were utilized
in the application part due to the absence of information regarding farmers’ initial
knowledge and experience in our dataset.
C.2.1.6 Expected Gain From Network: In the nal part of the simulation, based on the
values of the parameters we simulate the equation 8 and calculate the farmer’s
expected gain from learning through the network. e output of the simulation
captures the dierence between the farmer’s expected gain from three learning
channels (i.e., initial knowledge and experience, scouting, and learning from their
networks) and their expected gain as if they were not participating in the net-
work.
C.2.2 Simulating Value of Learning for the Base Network under Climate Change
Scenarios: For this part of the simulations, we repeat all the steps detailed in C.2.1
with two modications.
First, we modify step C.1.1 and dene three new distributions for the Growing Degree
Days by shiing the mean of the distribution by 10%, 20%, and 30%. en aer creating
a sample of 10,000 observations from each distribution, We focus only on the extreme
values of the distributions. erefore, we limit the growing degree days to the upper
tail of the distribution where all values are greater than two standard deviations from
the mean of the distributions.
42
Second, to calculate the expected gains from the networks by calculating a dierence-
in-dierence variable. First, we calculate the dierence between farmer’s expected
gain from the network and their expected gains as if they were not part of the network.
en, we calculate the same dierence under various climate change scenarios. Lastly,
we calculate the gap between the two calculated dierences. is gap captures the
farmer’s potential gain from the small network under each climate change scenario.
Firstly, we adjust step C.2.1.1 by dening three new distributions for the Growing
Degree Days, each with the mean shied by 10%, 20%, and 30% respectively. Sub-
sequently, we focus solely on the extreme values of these distributions. is entails
limiting the growing degree days to the upper tail of the distribution, where all values
surpass two standard deviations from the mean of the distributions.
Secondly, we modify the step C.2.1.6 to compute the expected gains from the net-
works, we dene a dierence-in-dierence variable. Initially, we calculate the gap
between a farmer’s expected gain from the network and their expected gains if they
were not part of the network. en, we repeat this calculation under various cli-
mate change scenarios. Finally, we determine the gap between the two calculated
dierences. e calculated gap captures the potential gain for farmers from the small
network under each climate change scenario.
e rest of the simulations remain unchanged from the instructions outlined in step
C.2.1.
C.3 Simulating the Value of Learning from Farmer’s Expanded Network
In this part, there are only two changes from section C.2 of this appendix :
C.3.1 First, we expand the farmer’s network by adding 5 new neighbors for each farm in
our baseline network. e new neighbors are added randomly within a 25-mile radius.
Using our data prepared in section C.1 of this appendix, rst, we randomly select ve
neighbors located within 25 miles of each farm. en we assign a unique identier
for those farms and using ArcGIS we connect those new neighbors to the network.
C.3.2 Number of iteration: Second, We modify the step C.2.1 of the previous section of
appendix C by changing the number of simulations to 4001. e rest of the simulations
remain unchanged. We follow exactly the same process as detailed in C.2.1.1 through
C.2.1.6.
We maintain the steps for simulating climate change scenarios unchanged from the
previous instructions outlined in step C.2.2.
C.4 Simulating the Value of Learning from Farmer’s Expanded Network with Signal
Selection
In this part of the simulation, we introduce criteria for signal senders. We assume that
farmers receive signals from the 10 nearest neighbors. is condition eliminates noisy
signals from farms located in far locations and improves the geographical relevance of the
signals.
In this part of the simulations, there are two deviations from the previous section C.3 of
the appendix C :
C.4.1 First, restrict the signal-sending farmers to be the only ten nearest farmers. e aver-
age distance for the signal-sending farm here is 6.8 miles with a minimum of approx-
imately 0 miles and a maximum of approximately 44 miles.
In our pool of signal-receiving farmers for the expanded network, we have 595 farms
and each receives 10 signals from their nearest neighbors.
C.4.2 Number of iteration: We set the number of simulations to 4001. en, we create a
loop to iterate through the following steps to simulate the value of the network. To
summarize our results on the graphs, we use dierent numbers of simulations to have
a similar frequency on the y-axis of the graphs. Since the number of observations
for our sub-samples varies when we limit our simulation for high and low-precision
signals, then we change the number of simulations for the low-precision sub-samples
to get a similar number of observations.
e other steps for simulating the value of learning from the network and the calcu-
lations under dierent climate change scenarios remain exactly the same as before.
43
Appendix D - Additional Simulation Results
Distribution of Gains for the Expanded SIRAC Network
(a) Scout precision: Low; Network precision: Low (b) Scout precision: Low; Network precision: High
(c) Scout precision: High; Network precision: Low (d) Scout precision: High; Network precision: High
Figure 5: Simulation of Farmer’s Expected Gains by Signal Precision - Expanded Network
Note: Figure 2 shows the distribution of farmer’s expected gain from learning from scouting and from the network
for the expanded SIRAC network with an application to management of ECB pest. Farmers have three channels
of learning: previous knowledge, scouting, and network. e blue histograms plot the distribution of farmer’s
gain from scouting relative to the reference case of only previous knowledge. e orange histograms plot the
distribution of farmer’s gain from scouting and networking relative to the case of only previous knowledge. e
dierence between the orange and blue histograms captures the gain from the network. e dashed vertical line
represents the median value of each distribution. Each graph plots distributions for dierent precision levels of
the signals from scouting and from the network.
44
Descriptive Statistics - Results
Table D.1: Descriptive Statistics - SIRAC Network
Variable obs P10 mean St.dev P90 Min Max
Signal Precision 47.6 × 10
6
0.43 0.59 0.13 0.85 0.29 1.00
Scouting Precision 47.6 × 10
6
0.57 0.59 0.12 0.86 0.14 1.00
Initial pest population 47.6 × 10
6
1.36 2.00 0.50 2.65 0.00 4.00
Pest Carrying Capacity 47.6 × 10
6
21.36 22.00 0.50 22.65 19.86 24.15
GDDs 47.6 × 10
6
1562 1900 302 2325 1500 3706
Distance (miles) 47.6 × 10
6
20.76 112.38 63.64 192.91 0.38 266.78
Corn Price 47.6 × 10
6
5.33 6.40 0.83 7.46 3.15 10.04
Average yield per acre 47.6 × 10
6
166.55 172.99 5.01 179.39 152.44 193.53
Expected Gain 47.6 × 10
6
$130 $298 $148 $494 $0 $635
Table D.1. summarizes the descriptive statistics of the parameters in our simulation model for the SIRAC
network and the farmer’s expected gain from the network, assuming that they receive ten signals from the
network.
45
Table D.2: Descriptive Statistics - Expanded Network
Variable obs P10 mean St.dev P90 Min Max
Signal Precision 23.8 × 10
7
0.42 0.61 0.20 0.89 0.43 1.00
Scouting Precision 23.8 × 10
7
0.57 0.75 0.17 0.93 0.14 1.00
Initial pest population 23.8 × 10
7
1.36 2.00 0.50 2.61 0.11 3.94
Pest Carrying Capacity 23.8 × 10
7
21.36 22.00 0.51 22.65 20.3 23.79
GDDs 23.8 × 10
7
1564 1903 301 2330 1500 3473
Distance (miles) 23.8 × 10
7
20.18 94.05 53.09 162.38 0.00 245.36
Corn Price 23.8 × 10
7
5.32 6.40 0.84 7.47 3.34 9.30
Average yield per acre 23.8 × 10
7
166.63 172.97 5.03 179.47 156.79 192.07
Expected Gain 23.8 × 10
7
$149 $348 $171 $576 $9 $649
Table D.2. summarizes the descriptive statistics of the parameters in our simulation model for the expanded
network and the farmer’s expected gain from the network, assuming that they receive ten signals from the
network.
46
Table D.3: Descriptive Statistics - Expanded Network with Signal Selection
Variable obs P10 mean St.dev P90 Min Max
Signal Precision 47.6 × 10
6
0.86 0.98 0.05 1.00 0.71 1.00
Scouting Precision 47.6 × 10
6
0.57 0.75 0.17 1.00 0.14 1.00
Initial pest population 47.6 × 10
6
1.36 2.00 0.50 2.65 0.02 3.94
Pest Carrying Capacity 47.6 × 10
6
21.36 22.00 0.50 22.64 19.91 24.11
GDDs 47.6 × 10
6
1561 1898 302 2324 1500 3692
Distance (miles) 47.6 × 10
6
2.23 6.81 4.12 9.74 0.00 24.70
Corn Price 47.6 × 10
6
5.33 6.40 0.83 7.46 2.97 9.65
Average yield per acre 47.6 × 10
6
166.64 173.044 4.97 179.42 151.057 193.49
Expected Gain 47.6 × 10
6
$223 $431 $188 $677 $5 $835
Table D.3. summarizes the descriptive statistics of the parameters in our simulation model for the expanded
network with signal selection and the farmer’s expected gain from the network, assuming that they receive 10
signals from the network.
47
Distribution of the Network Climate Adaptation Values for the SIRAC
Network
(a) Scout precision: Low; Network precision: Low (b) Scout precision: Low; Network precision: High
(c) Scout precision: High; Network precision: Low (d) Scout precision: High; Network precision: High
Figure 6: Simulation of Farmer’s Expected Gains under Climate Change by Signal Precision -
SIRAC Network
Note: Figure 4 illustrates the distribution of farmers’ expected gains from both the SIRAC network and scouting
activities under the scenario of a 10% increase in GDDs, specically focusing on the management of European
Corn Borer (ECB) pests. e blue histograms represent the distribution of dierences in farmers’ expected gains
from scouting alone, comparing the scenario aer a 10% increase in GDDs to the baseline scenario without climate
change. Meanwhile, the orange histograms show the distribution of dierences in farmers’ expected gains from
combining scouting and network signals, again comparing the post-10% GDD increase scenario to the no climate
change baseline. e dierence between the orange and blue histograms quanties the network’s adaptation
value under the scenario of a 10% GDD increase. is dierence highlights the additional benet that network
participation oers over scouting alone in adapting to climate change impacts. e dashed vertical lines in each
graph mark the median value of the distributions. Each subgraph within Figure 4 shows distributions for various
precision levels of scouting information and network signals.
48
