ArticlePDF Available

Abstract and Figures

We develop a model to evaluate the profitability of controlling rodent damage by placing barn owl nesting boxes in agricultural areas. The model incorporates the spatial patterns of barn owl predation pressure on rodents, and the impact of this predation pressure on nesting choices and agricultural output. We apply the model to data collected in Israel and find the installation of nesting boxes profitable. While this finding indicates that economic policy instruments to enhance the adoption of this biological control method are redundant, it does support stricter regulations on rodent control using rodenticides.
Content may be subject to copyright.
We develop a model to evaluate the profitability of controlling rodent damage by placing barn
owl nesting boxes in agricultural areas. The model incorporates the spatial patterns of barn owl
predation pressure on rodents, and the impact of this predation pressure on nesting choices and
agricultural output. We apply the model to data collected in Israel and find the installation of
nesting boxes profitable. While this finding indicates that economic policy instruments to enhance
the adoption of this biological control method are redundant, it does support stricter regulations
on rodent control using rodenticides.
Key words: agricultural damage control, environmental regulation, barn owl, rodent.
JEL codes: Q15, Q18, Q57.
Rodent damage to agriculture results in
double-digit percentages of yield reduction
across the globe (Singleton 2003;Leirs 2003).
This considerable damage suggests that the
effectiveness of conventional rodent-control
methods such as tillage, sanitation, trap-
ping, and rodenticide applications is limited
(Stenseth et al. 2003). The application of
rodenticides is frequently ineffective due
to the rapid immigration of rodents from
adjacent untreated areas, and because rodent
population outbreaks are unpredictable.
Moreover, rodenticides are often consid-
ered by farmers to be too costly (Skonhoft
et al. 2006;Davis et al. 2004;Stenseth et al.
2003). Risks of mortality by self-poisoning
Iddo Kan and Ayal Kimhi are affiliated with the Department
of Agricultural Economics and Management, The Hebrew
University of Jerusalem, Israel, and with the Center for
Agricultural Economic Research, Israel. Kimhi is also the
Deputy Director of the Taub Center for Social Policy Stud-
ies in Israel. Yoav Motro, Nir Horvitz, and Ran Nathan are
affiliated with the Department of Ecology, Evolution and
Behavior, The Hebrew University of Jerusalem, Israel. Yossi
Leshem and Yoram Yom-Tov are affiliated with the Depart-
ment of Zoology, The George S. Wise School of Life Sciences,
Tel Aviv University, Israel. We thank Shaul Aviel, Uria Sha-
hak, Myriam Freund, Michael Heiman, Shaul Ginzberg, Eyal
Lev, Yoav Cohen, Ruth Aviel, Neria Lifshitz, Dan Alon, Yael
Chassin, Michal Azaz, Ricky Ketner, Motti Charter, Uriel
Safriel, Rivka Rabinowitz, Gila Kahila Bar-Gal, Eitan Tcher-
nov, Moshe Coll, the editor, and two anonymous reviewers.
This study was partly funded by the USAid MERC, grant
TA-MOU-06-M25-078, and by the Center for Agricultural
Economic Research, Israel. Correspondence may be sent to:
(Eddleston 2000) and detrimental impacts
on non-target animals (Cox and Smith 1990)
are additional drawbacks of rodenticides.
The use of barn owls (Tyto alba) as a bio-
logical control method could be a more
cost-effective alternative, which might also
reduce the negative externalities associated
with rodenticides.
Barn owls, nocturnal raptors having a
nearly worldwide distribution, prey on a
variety of rodent species, many of which are
agricultural pests. Barn owls use pre-existing
cavities for nesting (Taylor 1994), and there-
fore face a scarcity in nesting sites. This trait
enables harnessing the barn owl’s hunting
abilities to control rodents in agricultural
fields by introducing artificial nesting boxes.
Nesting boxes are already used in various
parts of the world to help control a wide
range of crop-damaging rodents. For exam-
ple, such boxes are used to protect oil palms
in Malaysia (Duckett 1976) and rice in India
(Parshad 1999). However, there is insufficient
scientific information on how the technique
can be most effectively applied (Leshem
et al. 2010) and there is inconsistent evidence
of its damage-prevention success (Askham
1990;Wood and Fee 2003). Moreover, to the
best of our knowledge, the profitability of
rodent control using barn owls has never
been assessed. Consequently, there is a lack
of scientific knowledge that could prove
Amer. J. Agr. Econ. 96(3): 733–752; doi: 10.1093/ajae/aat097
Published online December 18, 2013
© The Author (2013). Published by Oxford University Press on behalf of the Agricultural and Applied Economics
Association. All rights reserved. For permissions, please e-mail:
by guest on May 22, 2014 from
734 April 2014 Amer. J. Agr. Econ.
useful to farmers who are planning rodent-
control activities and to regulators hoping to
spur adoption. This article contributes to the
literature on bioeconomics and agricultural
damage-control economics by conducting
a rigorous evaluation of the method’s prof-
itability. To this end, we develop a spatial
bioeconomic model and apply it using unique
agronomic and zoological data collected in
the agricultural fields of Kibbutz Sde Eliyahu,
Economic analyses of agroecological sys-
tems face conceptual and empirical chal-
lenges (Zhang et al. 2007), particularly due
to the presence of complex spatiotempo-
ral processes. Applications depend heavily
on the availability of biological and agro-
nomic data; examples include Brown, Lynch,
and Zilberman (2002),Nordblom et al.
(2002),Polasky et al. (2005),Griffiths et al.
(2008), and Polasky et al. (2011). Our data
enabled us to develop an economic model
that incorporates three functions associated
with spatial processes; the development of
these functions itself contributes to the scien-
tific disciplines of agronomy, biogeography,
and animal movement behavior. The first
function describes barn owls’ spatial preda-
tion patterns. During the breeding season
(March–October), the nesting place con-
stitutes a point source for the barn owl’s
predation activity.1Hence, the predation
pressure exercised by barn owls on their
surroundings is expected to diminish with
distance from the nesting place. Behavioral
examination of other raptors has revealed
variability in predation attractiveness across
land uses (Thirgood, Redpath, and Graham
2003). Using radiotelemetry records of barn
owl locations, this study is, to our knowledge,
the first to estimate a function of spatial dis-
tribution of barn owl predation pressure that
accounts for the impact of both distance and
land-use appeal. The second function models
the selection of nesting boxes by barn owls
for breeding. Previous estimates indicate
that the probability of nest-box occupancy
depends on box attributes (e.g., the entrance
aspect) and on its location in relation to
various land uses (Frey et al. 2011;Charter
et al. 2012). Here, we estimate a nest-box
1A barn owl’s predatory act starts and ends at its nesting place,
and the bird catches a single prey in each hunting act (Lessells
and Stephens 1983). These habits make the nesting place the
point source of the barn owl’s spatial impact on rodents.
occupancy probability function that explic-
itly incorporates the impact of a wide range
of crops, the distance between one nesting
box and others, and the occupancy status
of the box in the previous season. Taking
advantage of the spatial variability of barn
owl predation pressures on fields, our third
(crop-production) function treats the barn
owl predation pressure as damage-control
An additional contribution of this study
stems from the recursive estimation pro-
cess we use, which ensures consistency of
the biospatial processes associated with the
three functions: the estimated predation-
pressure function is used for computing
predation-pressure variables that serve as
explanatory variables in the estimation of
both the box-occupancy probability function
and the crop-production function. As a result,
we obtain an integrative functional system
that enables us to evaluate the contribution
of barn owls to agricultural outputs through
simulations of nesting-box locations. The
locations of the boxes in relation to land
uses impact field outputs by determining the
expected predation pressures applied on the
fields, where these expected pressures are
the products of the predation pressures exer-
cised on the fields from occupied boxes, and
the probability of the boxes being occupied.
Specifically, we use the model to compute
alfalfa (Medicago sativa) outputs under three
scenarios of nesting-box distribution: (a)
under the observed locations of the 58 boxes
currently placed in the Kibbutz’s fields, (b)
in the absence of these boxes, and (c) under
a simulated distribution of the 58 boxes that
maximizes alfalfa-production profits. These
three simulations enable us to evaluate both
the contribution of the boxes to profit in their
current locations and the extent to which
profits could potentially be increased.
Our dataset is a panel of detailed zoo-
logical and agronomic information that is
unique in its suitability to our study. First,
barn owl nesting boxes have been placed in
the Kibbutz’s fields since 1983 (Motro et al.
2010), so that during the period covered by
our data (1999–2008), the barn owl popula-
tion was already familiar with the presence
of the boxes.2Second, the area covered is
2According to Wood and Fee (2003), the presence of nesting
boxes may also affect the barn owl population, and thus the
occupancy probability of boxes and rodent-damage control. We
controlled for this potential dynamic effect in the estimation of
by guest on May 22, 2014 from
Kan et al. Agricultural Rodent Control Using Barn Owls: Is it Profitable? 735
sufficiently large to encompass the sizable
spatial range influenced by barn owls. Third,
the agricultural lands are heterogeneous, so
that our analysis can account for the impact
of different land uses on the spatial patterns
of barn owl predation pressure, and on the
selection of nesting places. Finally, agricul-
tural production is centrally managed by
the Kibbutz, so that variability in skills, con-
straints, and other management factors is
Barn owls affect agricultural outputs only
indirectly through their predation impact on
rodents, which in turn affect crops through
herbivory. The impact of rodent herbivory is
spatially distributed in relation to rodents’
breeding places. Thus, a comprehensive
analysis should account for the spatial dis-
tributions of both the raptors and their prey.
Unfortunately, due to high monitoring costs
and methodological difficulties, reliable
and continuous estimates of rodent spatial
distribution and population size were not
obtainable. Therefore, our analysis over-
looks the explicit process by which rodents
channel the impact of barn owls on yield.
To separate the indirect damage-control
effect of the barn owls from the direct impact
of production inputs, we adopt the formu-
lation approach used by Lichtenberg and
Zilberman (1986),Babcock, Lichtenberg, and
Zilberman (1992),Blackwell and Pagoulatos
(1992), and others, wherein a damage-
abatement function is integrated into the
production function. Specifically, we apply
the model developed by Saha, Shumway,
and Havenner (1997), which also allows
controlling for the interactive effects of pro-
duction inputs, damage-abatement inputs, and
external factors.
Due to data availability, we focus on
alfalfa production, which is a perennial,
multiharvested legume grown mainly for
fodder. Alfalfa is highly prone to rodent
damage because rodents accumulate over
the multiannual crop growth period, while
almost no agromechanical measures can
be implemented in the fields (Moran and
Keidar 1993). Poisoning rodents in alfalfa
exhibits low performance due to the con-
stant presence of the fresh, nutritious, green
foliage favored by the rodents (Proulx
1998). This makes alfalfa a good case study
the box-occupancy probability function, but found no significant
for examining the economic efficiency of
agricultural rodent control by using barn
Our results indicate that barn owls’ con-
tribution to the Kibbutz’s alfalfa outputs,
under the observed locations of the nest-
ing boxes, amounts to nearly 10%, yielding
a net-profit increase of more than $200/ha
per year. Given that we overlook potential
contributions to the outputs of other crops,
our evaluation probably underestimates
the overall profit contribution of barn owls.
Note that our evaluation refers to the profit
contribution of the barn owls nesting in the
Kibbutz’s nesting boxes only, and it measures
the profit contribution above and beyond the
rodent-control effects of agronomic activ-
ities, sanitation, and farmer-independent
factors within and outside of the studied area
(e.g., barn owls and other raptors nesting in
unmonitored places). However, as farmers
at Sde Eliyahu completely avoid rodenti-
cide application for ideological reasons, our
dataset cannot be used to directly compare
biological and conventional rodent-control
The Model
Our model describes the management of a
nesting box system subdivided into Ifields in
a farm area, where all other agrobiological
impacts are considered exogenous. Since our
analysis relies on the spatial distribution of
predation pressure, intra-field variations need
to be captured; therefore, we consider an
artificial division of field i,i=1, ...,I, into
Miuniform-sized land units (e.g., hectares).
Let mi,mi=1, ...,Mi, denote a specific land
unit in field i. The unit miis geographically
represented by its central point, the coor-
dinates of which are incorporated in the
two-dimensional column vector umi. The two-
row matrix ui=um1,...,uMiincorporates
the coordinates of all land units in field i,
and u=(u1,...,uI)is the coordinate matrix
of all land units on the farm. The vector vi
contains all other specific attributes of field i,
such as soil type, microclimate, and installed
infrastructures, and v=(v1,...,vI)is defined
Spread over the farm are Knesting boxes,
where xkis the two-dimensional coordi-
nate vector of the location of nesting box
k;k=1, ...,K; and x=(x1,...,xK)is the
matrix of the coordinates of all Kboxes. The
by guest on May 22, 2014 from
736 April 2014 Amer. J. Agr. Econ.
suitability of box kas a nest is affected by
a vector of exogenous attributes, ak, such as
shading conditions, and by a set of features
of the box, denoted hk, including the box’s
height, color, and entrance direction. The
matrices a=(a1,...,aK)and h=(h1,...,hK)
are defined accordingly.
Various crops, j, are routinely grown on the
farm. Let δij be an indicator variable that has
a value of 1 if crop j,j=1, ...,J, is assigned
to a specific field, i.TheI×Jmatrix of crop
attributions to fields, δ, is defined accordingly.
The function
rij (u,x,δ,a,h)=
likj (dik (ui,xk))
represents the cumulative predation pres-
sure applied by barn owls nesting in the K
boxes on the rodents within Miland units
of field i, when this field is allotted to crop j.
Cumulative pressure is defined as the sum
of the products of two functions. The first
function in equation (1), likj (dik (ui,x)),is
called the predation-pressure function, and
represents the intensity of the predation
activity applied by a barn owl nesting in box
kin the area of field iif the field is devoted
to crop j. This intensity is measured by the
probability that the barn owl will hunt in field
i, which in turn is a function of dik (ui,xk),
the vector of distances from box kto all Mi
area units of field i. The predation-pressure
function is crop-specific, and therefore differ-
entiates the hunting preferences of barn owls
toward different crops. The second function
in equation (1), Sk(u,x,δ,ak,hk), is termed
the box-selection function, and expresses
the probability of box kbeing selected as
a nesting place by a pair of barn owls. This
probability depends on the box’s attributes
akand hk; it also depends on the features of
the box’s surrounding environment, includ-
ing the type of crops allotted to the farm’s
fields, represented by δ, and the distances to
all other boxes and land units in the farm,
which in turn depend on the locations of all
the boxes (x) and land units (u). The land
use around the box may affect its occupancy
rate if barn owls prefer nesting in boxes close
to land uses that are favored for predation.
The relevance of the distance to other boxes
stems from the potential territorial behavior
of barn owls. In our empirical estimation of
the box-selection function, we control for
these spatial effects using variables, which,
for consistency, were calculated based on the
estimated predation-pressure function.
Let qij be the vector of production inputs
applied to field iwhen it is assigned to crop
j, and let matrix qincorporate the I×Jsets
of qij production inputs. The vectors b1and
b2incorporate exogenous factors associ-
ated with production and costs, respectively,
including managerial skills, climate condi-
tions, input constraints, and prices of inputs
and outputs. Given u,a,v,b1, and b2, a profit-
maximizing farmer will decide on the optimal
assignment of crops to the fields, δ, the appli-
cation of inputs to the fields, q
ij,i=1, ...,I,
j=1, ...,J, the features of the boxes, h
k=1, ...,K, and the location of the boxes, x.
The maximal profit is:
ijpjHij (rij(u,x,δ,a,h),(2)
where pjis crop j’s output price,
Hij(rij (u,x,δ,a,h),q
ij,vi,b1)is a production
function specific to field iand crop j, and c(·)
is a farm-level cost function.
Equation (2) integrates the manage-
ment tools and the spatial agrobiological
relationships into one comprehensive
function, which constitutes the basis for
our empirical analysis. Note that our
estimation of the production function
Hij(rij (u,x,δ,a,h),q
ij,vi,b1)uses the model
developed by Saha, Shumway, and Havenner
(1997), wherein the field-level inputs q, field
characteristics v, and farm-level attributes
b1are all associated with direct production
impact, as well as indirect effects through
damage control. In addition, while decid-
ing on the assignment of crops to fields, δ,
farmers may account for the box locations, x.
However, in our empirical analysis we found
that the observed assignment δis indepen-
dent of x; that is, δis dictated primarily by
other considerations, such as crop rotation,
and is therefore exogenous.
Empirical Specifications and Estimation
Our empirical application uses data from
1999–2008, covering the lands of Kibbutz
by guest on May 22, 2014 from
Kan et al. Agricultural Rodent Control Using Barn Owls: Is it Profitable? 737
Sde Eliyahu in the Beit She’an Valley, Israel
(3230N, 3530E). This is a semi-arid region
with average annual rainfall of about 250 mm,
mild winters, and dry, hot summers. Fifty-
eight barn owl nesting boxes were placed in
various fields between 1983 and 1996. The
12.5 km2study area comprises heterogeneous
land uses—altogether 11(J=11)—including
residential zones, field crops, and fruit plan-
tations. There were 54 fields (I=54), which
we divided into land units (mi)of56m
each. The spatial distribution of boxes across
diverse land uses enabled us to investigate
how the variation in land use affects the
predation pressure exerted on rodents by
barn owls, as well as the occupancy rates of
nesting boxes. In the following sub-section,
we apply a survival analysis to estimate a
function describing the spatial distribution
of the predation pressure exercised by the
barn owls on the various land uses around
the nesting boxes. This function generates
explanatory variables that we used to esti-
mate both the box-selection function and the
alfalfa-production function.
The Predation-Pressure Function
The predation-pressure function likj(dik (ui,
xk)) is defined as the probability that a barn
owl nesting in box kwill search for prey in
field iif the field is devoted to crop j,asa
function of the distance between the box and
the field. A field located further from the nest
is expected to be less appealing to the owls
because longer flights are required. The
appeal of the field also depends on the crop
grown there, through the size of the rodent
population it attracts, and the preying condi-
tions it offers. For instance, the presence of
perching points and the heights of plants may
affect the barn owl’s chances of pinpointing
its prey, as well as its hunting success.
Let dmik(umi,xk)be the distance between
box kand the centroid of land unit miwithin
field i. The predation-pressure function for
field i, when assigned to crop j, is:
(3) likj(dik (ui,xk)) =
where Prj(dmik(umi,xk)) is a probability
density function specific to crop j, which
represents the probability of the barn owl
nesting in box kto hunt at point mi.
Estimating Prj(dmik(umi,xk)) requires two
datasets. The first represents the barn owl’s
“actual predation behavior,” and includes
records of the locations of barn owls while
hunting, from which the crop at each hunting
location can be identified and each bird’s dis-
tance from its nesting box can be computed.
The second dataset reflects the “predation
opportunities” of barn owls, and is specifically
required to estimate differences in the hunt-
ing appeal of different crops. We employed a
recursive estimation procedure that involves
survival analyses based on radiotelemetry
records of barn owls in Sde Eliyahu’s fields,
the selection of a functional form based on
the Akaike Information Criterion (AIC), and
estimation of crop-preference parameters. In
the online appendix A, we detail the data and
the estimation procedure. Here we report
the estimation result of the selected gamma
probability density function:
0.11 exp 0.16 ln πd2
0.025 exp 6.35 ln πd2
where ajis a crop-preference parameter
whose value is presented in table 1for the
11 land uses. Except for the case of legumes,
barn owls seem to prefer trees, probably
owing to the advantage provided by perches
(Kay et al. 1994).
The estimated predation-pressure func-
tion has a few noteworthy implications.
First, our evaluation of the contribution of
barn owls to agricultural profits is based on
alfalfa production. Given that the estimated
crop-preference parameters for alfalfa are
among the lowest, this crop is not attractive
to barn owls, and therefore our focus on
alfalfa may considerably underestimate the
overall contribution of barn owls to profit.
Second, as our evaluations rely heavily on
the predation-pressure function, validation is
required. In the on-line appendix B, we pro-
vide empirical evidence showing that (a) the
calculated predation pressures can explain
the hunting patterns of barn owls in alfalfa
fields as found in the radiotelemetry survey,
and (b) endogeneity due to reverse causality
is unlikely, meaning that larger alfalfa yields
do not stimulate predation pressure. Finally,
the predation-pressure function is also used
by guest on May 22, 2014 from
738 April 2014 Amer. J. Agr. Econ.
Table 1. Estimates of Crop-Preference Parameters
Land Use αj(tvalue) R2F-statistic
Fallow 0.69 (3.16)∗∗∗ 0.40 9.96
Alfalfa, year 1 0.39 (2.54)∗∗ 0.30 6.46
Alfalfa, year 2+0.29 (2.51)∗∗ 0.30 6.29
Corn 0.12 (1.95)∗∗∗ 0.20 3.80
Legumes 2.08 (4.66)∗∗∗ 0.59 21.74
Wheat 0.79 (3.12)∗∗∗ 0.39 9.75
Vegetables 0.42 (2.49)∗∗ 0.29 6.22
Citrus 1.26 (2.76)∗∗ 0.34 7.63
Dates 2.64 (5.71)∗∗∗ 0.69 32.65
Olives 1.89 (1.90)0.19 3.61
Residential areas 0.88 (4.91)∗∗∗ 0.62 24.11
Note:denotes significance at a 10% level, ∗∗ denotes significance at a 5% level, and ∗∗∗ denotes significance at a 1% level.
to validate our assumption that the box loca-
tions (x) do not affect the optimal assignment
of fields to crops (δ).3
The Box-Selection Function
From the farmer’s point of view, the prof-
itability associated with installing and
maintaining a nesting box depends primarily
on the probability of the box being in use.
The proximity of the box to attractive hunt-
ing areas has a potentially important impact
on the probability of the box being occupied.
Hence, the location of each box is likely to be
a key determinant of its occupancy rate, and
in turn a key determinant of the efficacy of
rodent control by barn owls overall.
Barn owls reselect their nesting places
once a year, at the onset of the breeding
season. Previous studies have indicated the
dependence of nesting rate in boxes on their
physical features and geographical attributes
(Charter et al. 2010). In this study, we esti-
mate an occupancy probability function
in which the impacts of the boxes’ prox-
imity to certain crops and to other boxes
are incorporated based on the estimated
predation-pressure function.
The average occupancy rate for the 58
boxes over the 10-year period is 43%, ranging
from 20% to 62%. For comparison, Wood
and Fee (2003) report occupancy rates of
70% in oil palm estates in Malaysia. Our
explanatory variables can be classified into
3We estimated a multinomial logit model of the probability that
fields are assigned to non-perennial crops, in which the calculated
cumulative predation pressure exercised by all of the boxes on
the fields constitutes the explanatory variable; the corresponding
coefficients were found to be statistically insignificant for all crops.
three groups. The first group includes time-
invariant features of the boxes themselves,
including dummy variables for three entrance
directions and a dummy for boxes located
in the shade. The second group of variables
represents the environment of each box. We
hypothesize that boxes located closer to land
uses that provide better hunting conditions
are more attractive for nesting. Variables
in this group represent the predation pres-
sure exerted by barn owls nesting in a box
on rodents in the aforementioned 11 land
uses (see table 1), conditional on its being
occupied. The predation pressure was com-
puted for each box k, land-use j, and year t,
t=1, ..., 10, by:
(5) Lkjt =
δijt likj(dik (ui,xk))
using the parameters estimated for the
predation-pressure function likj(dik (ui,xk)).
The predation pressure represents the prob-
ability (or fraction of time) that a barn
owl nesting in box kwill hunt in the fields
assigned to crop j. Note that the pressures
computed for Sde Eliyahu’s residential area
and for perennial plantation fields are time-
invariant, since δijt =δij for all t=1, ...,10.
In addition, to control for potential terri-
torial effects in nest selection, we include
an “engagement probability” variable
that measures the probability of interac-
tion between a barn owl nesting in box k
and those nesting in all other boxes. The
engagement-probability variable is calcu-
lated by applying equation (4) to the distance
from box kto every other box, and averaging
across boxes. For all of the variables in this
by guest on May 22, 2014 from
Kan et al. Agricultural Rodent Control Using Barn Owls: Is it Profitable? 739
group (i.e., the predation pressures on the 11
land uses and the engagement probability),
second-degree polynomials are included to
allow for non-linear effects.
It is worth noting that yields in the fields
surrounding the nest might affect the nest-
ing rate, and controlling for this effect could
introduce endogeneity into our analysis.
However, this is not relevant for the spe-
cific case of alfalfa, since box occupancy is
generally established before the first alfalfa
The third group of variables includes
time-specific variables—annual rainfall at
Sde Eliyahu, the number of years since the
box was installed, year fixed effects, and the
total number of nesting boxes throughout
Israel to control for potential impacts on
the barn owl population. In addition, since
boxes occupied in previous years may signal
favorable nesting conditions, we included
a lagged dependent variable that indicates
whether the box was occupied in the previous
year. This entails estimating a dynamic probit
model, with unobserved box heterogeneity.
Following Wooldridge (2005), we internalize
into the model the correlation between the
initial dependent variable (denoted gk0) and
the unobserved heterogeneity (ηk)using a
linear function: ηk=θ0+θ1gk0+zkθ2+φk,
where zk=(zk1,...,zk10)is the row vec-
tor of all explanatory variables in all time
periods, θ0,θ1, and θ2are coefficients, and
φk|(gk0,zk)N(0, σ2
φ). This yields a dynamic
probit model with response probability:
Pr(gkr =1)=(zkt ψ+ζgk,t1+θ0
where gkt is the dichotomous dependent vari-
able, zkt is a vector of exogenous variables,
and ψand ζare the coefficients of interest.
This equation can be estimated by standard
random-effect probit software (e.g., by the
xtprobit command in Stata).
Due to sample size limitations, the zk
vector in our application incorporates, for
each year t, the sum of the predation pres-
sures over all non-perennial crops, which
are the time-variant variables. The stepAIC
procedure (Venables and Ripley 2002) was
employed to select the set of variables to
be retained in the model based on the AIC.
Table 2reports the estimation results.
Only one variable from the group of
time-invariant features of the boxes was
retained by the stepAIC procedure: the
shade conditions.4As could be expected
in hot environments, shaded boxes are sig-
nificantly more attractive. The agricultural
environment also appears to play an impor-
tant role in nesting box occupancy. The
coefficients of alfalfa and wheat, which are
known to be favored by rodents, are pos-
itive. That is, larger predation pressure on
these crops (i.e., a reduction in the distance
between the box and the fields where these
crops are grown, which in turn increases
the probability that the barn owl nesting
in the box will hunt in these fields) increases
the probability of the box being occupied.
However, this finding is inconsistent with the
predation habits (table 1), implying that hunt-
ing patterns might not always be accounted
for in the nest-selection stage. Note that
the variable measuring the predation pres-
sure on alfalfa fields in their first production
year was eliminated by the stepAIC pro-
cedure; this indicates a possible learning
process in which barn owls gradually recog-
nize the alfalfa fields, or come to understand
their appeal. Proximity to date palms also
increases nesting probability,5possibly due
to the preference for perches as prowl points
and the abundance of rodents in date plan-
tations, particularly rats.6On the other hand,
barn owls tend to avoid nesting in boxes
located close to residential areas. This may
be attributed to territorial effects, as some
barn owls routinely nest in Sde Eliyahu’s
residential areas, or to aversion to human
presence, light, and noise. We hypothesize
that the distance between neighboring nests
will reflect an attraction-repulsion balance,
converging to some favorable intermediate
distance.7This hypothesis is reinforced by
the opposite signs of the coefficients of the
engagement-probability and engagement-
probability-squared variables, which indicate
4The dummy variable indicating the east-facing side of the box
entrance was retained in the second-best model of the stepAIC
procedure, which has a relative likelihood that is 7% lower than
the first best (Burnham and Anderson 2002).
5The marginal effects of the crops’ predation-pressure variables
with statistically significant non-linear effects (pressures on dates
and residential areas) are found to be monotonic throughout the
whole sample range of these variables.
6Shaul Aviel, personal communication, Sde Eliyahu, March
7Barn owls exhibit type B territorial behavior patterns (sensu
Taylor 1994), where the area of breeding activity (the nest) is
defended, but the hunting area is not. This implies that the
territorial effect will be limited, and even reversed, if distances
between boxes become large enough, and barn owls may tend
to avoid nesting in boxes that are too isolated.
by guest on May 22, 2014 from
740 April 2014 Amer. J. Agr. Econ.
Table 2. Estimation Results for the Dynamic Probit Nesting Function
Sample Mean Coefficient Marginal Probability
Variable (St. Dev.) (Zvalue) Effect (tvalue)a
Occupancy (gkt, dependent
0.4293 (0.4954)
Shaded conditions
0.2414 (0.4283) 0.582 (2.74)∗∗∗ 0.239 (2.47)∗∗
Pressure on alfalfa, year 2+
0.0302 (0.0499) 2.452 (1.84)1.009 (1.84)
Pressure on wheat
0.1014 (0.0933) 1.696 (1.83)0.698 (1.67)
Pressure on dates
0.0407 (0.0651) 11.42 (2.20)∗∗ 3.074 (1.70)
Pressure on dates squared 0.0059 (0.0127) 48.61 (2.09)∗∗
Pressure on residential
areas (Probability)
0.0105 (0.0150) 87.48 (3.16)∗∗∗ 24.95 (1.93)
Pressure on residential
areas squared
0.0003 (0.0008) 1,278 (3.04)∗∗∗
Engagement probability
0.1232 (0.0683) 7.759 (2.04)∗∗ 0.211 (0.38)
Engagement probability
0.0198 (0.0246) 29.40 (2.62)∗∗∗
Annual rainfall (cm/year) 24.780 (7.973) 0.047 (3.48)∗∗∗ 0.019 (4.11)∗∗∗
2003 (dummy) 0.1000 (0.3003) 0.501 (1.57) 0.206 (1.83)
2004 (dummy) 0.1000 (0.3003) 0.653 (3.26)∗∗∗ 0.269 (2.81)∗∗∗
Occupied in previous year
(gk,t1) (dummy)
0.3931 (0.4889) 0.793 (6.31)∗∗∗ 0.327 (5.36)∗∗∗
Occupied in 1998 (gk0)
0.1379 (0.3451) 0.535 (2.77)∗∗∗ 0.220 (3.02)∗∗∗
Sum of pressures on
variant crops in 2002
0.3148 (0.1879) 8.496 (1.42) 3.497 (1.04)
Sum of pressures on
variant crops in 2004
0.3223 (0.1873) 11.55 (1.60) 4.755 (1.25)
Sum of pressures on
variant crops in 2007
0.3100 (0.1906) 4.987 (1.67)2.053 (1.37)
Constant 1.725 (4.24)∗∗∗
σφ3.49 ×104
φ)11.22 ×107
Observations 580
Log likelihood 306.5
AIC 649.1
Pseudo R20.23
Notes: denotes significance at a 10% level, ∗∗ denotes significance at a 5% level, and ∗∗∗ denotes significance at a 1% level.
aThe marginal effects were evaluated for the mean values of the explanatory variables. Standard errors of the marginal effect statistics were calcu-
lated by bootstrap procedure.
the existence of a distance between boxes
at which occupancy rate is maximized. By
employing the estimated probability den-
sity function (equation 4), we found that,
ceteris paribus, an average distance of 410m
between a box and all other boxes maximizes
the box’s occupancy probability. For compar-
ison, the actual average distance between a
box and all other boxes in the study area was
approximately 1 km. In other words, increas-
ing the density of the 58 boxes might increase
their occupancy rates.
The coefficient of the annual rainfall vari-
able is positive. This can be explained by the
associated higher availability of vegetation
as food in the fields and waterways, which
in turn stimulates the population growth of
rodents and possibly other prey.
by guest on May 22, 2014 from
Kan et al. Agricultural Rodent Control Using Barn Owls: Is it Profitable? 741
Occupancy in the previous year (the lagged
dependent variable gk,t1) has a significant
positive effect, indicating the potential impor-
tance of signals that might be maintained
in the boxes between years.8There is also
a strong correlation between the unob-
served heterogeneity (ηk) and the initial
value of the dependent variable (occupied in
1998, gk0). On the other hand, ηkis weakly
correlated with the sums of the predation
pressures on time-variant crops that were
retained by the stepAIC procedure. Employ-
ing a log-likelihood test, the hypothesis
φ1=0 was not rejected, imply-
ing that the panel probit estimator does not
significantly differ from the pooled probit
The Alfalfa-Production Function
Alfalfa is routinely grown at Sde Eliyahu.
Each year, on average, 42 of the total 540ha
of agricultural land at Kibbutz Sde Eliyahu
are allocated to alfalfa. Our data encompass
a panel of 429 alfalfa harvests in 21 fields
from 1999–2008. An alfalfa field is cultivated
and sown during the autumn of the first pro-
duction year, remains untreated during the
rainy winter season, and is then harvested
up to 11 times during the dry seasons from
April to September for a few years—usually
not more than 4 sequential ones due to yield
reduction. The fields are fertilized once every
autumn, and irrigated twice following each
harvest. The irrigation dose per harvest is
determined according to the growing period
only, and is therefore exogenous.9Most fields
are irrigated by sprinklers, and some by a
moving platform, which enables treatment
against rodents by flood irrigation. Rodent
control in all fields is based on barn owls
and other factors, including other predators,
natural flooding of canals during the winter,
and agronomic activities such as plowing,
control of other pests, and routine sanitation
of field margins and waterways. Nevertheless,
as reflected by the high proportion of rodents
in the barn owls’ diet (Charter et al. 2009),
a large rodent population is present in Sde
Eliyahu’s fields.
8While mature barn owls tend to stay year-round within a
certain region, in our sample they rarely returned to the same
nesting box in a subsequent year.
9Rainfall events during the harvest seasons are rare, and the
irrigation dose of each harvest corresponds to the months of its
growing period: 140 mm/harvest in April, May and September,
160 mm/harvest in June, and 180mm/harvest in July and August.
To estimate the alfalfa-production func-
tion, we adopt the model developed by Saha,
Shumway, and Havenner (1997):
(7) y=H[W,G(X,Q)]
where yis the quantity produced, Wis a vec-
tor of direct production inputs, Xdenotes a
vector of damage-control agents, G(·)is the
abatement function, and QWis a subset
of the inputs in Wthat, in addition to their
direct impact, also indirectly affect the yield
through interactions with damage-control
agents. For example, irrigation may directly
increase yields and at the same time change
the rodent population, thereby altering
the effectiveness of rodent control by barn
owls. Saha, Shumway, and Havenner (1997)
suggested the empirical specification:
(8) y=Y(W,β)·G(X,Q,ω,e)·exp(ε)
(9) G(X,Q,ω,e)=exp [A(X,Q,ω)e]
is the abatement function, Y(·)and A(·)are
continuous and differentiable functions, β
and ωare vectors of parameters, and eand
εare error terms. By assuming eN(μ,1
εN(0, 1)and cov(e,ε)ρ, one obtains the
heteroscedastic error term εA(·)e. Hence:
(10) lnyN[ln Y(·)μA(·),B(·)]
where B(·)1+A(·)22A(·)ρ. These
tractable assumptions allow an exact for-
mulation of the expectation of output and its
y=Y(·)·exp B(·)
Following Saha, Shumway, and Havenner
(1997), we assume a linear specification for
(13) A(X,Q,ω)=ω0+ωXX+ωQQ
and a Cobb-Douglas functional form for Y(·):
(14) Y(W,β)=
by guest on May 22, 2014 from
742 April 2014 Amer. J. Agr. Econ.
where βςis the coefficient of the continuous
variable Wς,WDis a set of dummy vari-
ables, and βDis their corresponding vector of
Our dependent variable is yi1th, the weight
(in ton/ha) of alfalfa harvested from field
iduring harvest number hin year t, where
j=1 denotes alfalfa production. The set of
damage-control variables in Xincludes only
the effect of the barn owls, represented by
the cumulative predation pressure exer-
cised by the occupied nesting boxes in the
alfalfa fields. For some field ithat is assigned
to alfalfa (j=1) in year t, this cumulative
pressure represents the probability that a
barn owl nesting in some box kwill hunt in
this field, summed across all of the occupied
boxes in Sde Eliyahu in that year. In view of
equation (1), the cumulative probability is
calculated for each field i, which is assigned
to alfalfa in year tby:
(15) ri1t=
ξkt lik1(dik (ui,xk))
where ξkt is an indicator variable receiving a
value of 1 if box kis occupied in year t, and
0 otherwise. The per-hectare cumulative pre-
dation pressure of the field is included in X,
as well as its square, to control for a potential
non-linear effect.
The vector Qincorporates all of the other
explanatory variables, encompassing the
vectors q,v, and b1in equation (2). These
variables include rainfall (mm/year), irriga-
tion (mm/harvest), field size (ha), time since
the previous harvest (days), average temper-
ature during the period from the preceding
harvest, and a set of dummy variables indicat-
ing the availability of flood irrigation in the
field, the assignment of the field to organic
production, the production year (1999 to
2008), the field (21 fields), the serial year of
production in the field (ranging from 1 to 4),
and the serial harvest number (ranging from
1 to 11). We assume that all of these explana-
tory variables are included in Q(i.e., Q=W)
for two reasons: first, fields under conven-
tional production obtain similar treatments
against pests and herbs before the harvesting
period. That is, other than barn owl predation
pressure, there is no variability across these
alfalfa fields with respect to damage-control
variables. Organic production entails the
avoidance of not only pesticides, but also
fertilization; therefore, the dummy variable
indicating organic production cannot be
excluded from W. The second reason for our
assumption stems from the size of our set
of explanatory variables, which is too large
relative to the sample size to detach variables
from Winto a Qsubgroup based on a sep-
arability pretest (see Saha, Shumway, and
Havenner 1997).
The parameters ω,β,μ, and ρwere esti-
mated by maximizing the log-likelihood
function (LLF):
+[lnyi1th(·)ln Yi1th (·)+μAi1th(·)]2
using a non-linear maximization technique.
The hypothesis that the error term εA(·)e
is normally distributed was not rejected
(P-value =0.22). Table 3reports the sam-
ple means and standard deviations of the
variables, the estimated coefficients of the
functions Y(·)and A(·), and the marginal
effects of the variables on production mean
and variance.
The formulation in equation (9) implies
that the effect of each variable in A(·)on
damage abetment is opposite in sign to its
estimated coefficient. The marginal effect of
the predation pressure is positive, implying
that barn owls abate damage (i.e., the pro-
duction of an alfalfa field increases with the
probability that barn owls will hunt in that
field, where the increase is channeled by the
damage-abatement element of the produc-
tion function). This finding provides evidence
for a real contribution of the barn owls to
agricultural productivity. The contribution of
the predation pressure exhibits a diminishing
return to scale, as the coefficients of preda-
tion pressure and predation-pressure squared
have opposite signs. However, since the
coefficient of the squared variable is not sta-
tistically significant, we cannot reject convex
responses of abatement to predation-pressure
10 Nevertheless, even a concave response of the output to a
change in the level of damage control can stem from various
characteristics of the damage-abatement process; therefore, as
shown by Foxand Weersink (1995),we cannot make any deductions
from these findings on the nature of the impact of barn owls
on rodents, or on the relations between rodent populations and
by guest on May 22, 2014 from
Kan et al. Agricultural Rodent Control Using Barn Owls: Is it Profitable? 743
Table 3. Estimation Results for the Production and Damage-Control Functions
Coefficient (Zvalue) Marginal Effect (tvalue)a
VariablebSample Mean (St. Dev.) Production, Y(·)Damage Control, A(·)Mean E(y) Variance V(y)
Production (dependent
variable, ton/harvest-ha)
1.910 (0.643)
Predation pressure
0.028 (0.038) 0.735 (2.66)∗∗∗ 123.4 (5.64)∗∗∗ 42.266 (3.15)∗∗∗
Predation pressure squared 2.24 ×104(4.55 ×104) 23.49 (1.54)
Flood irrigation (dummy) 0.443 (0.497) 1.022 (1.95)5.873 ×103(1.96)0.472 (0.43) 0.051 (0.05)
Irrigation (mm/harvest) 139.1 (52.31) 0.017 (3.17)∗∗∗ 2.602×106(0.36) 0.003 (1.16) 9.314 ×104(2.60)∗∗
Organic (dummy) 0.387 (0.488) 0.591 (3.00)∗∗∗ 4.183 ×103(2.89)∗∗∗ 0.062 (0.08) 0.143 (0.31)
Growing period
22.26 (10.51) 0.026 (3.65)∗∗∗ 1.815 ×104(4.70)∗∗∗ 0.011 (0.50) 3.75 ×103(0.76)
Temperature (C) 31.52 (4.522) 0.105 (0.68) 1.959 ×105(0.16) 0.014 (0.68) 4.87 ×103(1.08)
Precipitation (mm/year) 248.9 (65.60) 0.059 (0.90) 4.585 ×105(0.08) 0.01 (0.58) 3.31 ×103(1.87)
Field size (ha) 10.37 (4.305) 0.257 (2.09)∗∗ 3.082 ×104(1.89)3.01 ×104(0.59) 1.02 ×105(0.98)
Year no. 2 (dummy) 0.445 (0.498) 0.630 (6.82)∗∗∗ 4.591 ×103(8.93)∗∗∗ 0.103 (0.10) 0.110 (0.41)
Year no. 3 (dummy) 0.138 (0.345) 0.267 (2.75)∗∗∗ 7.861 ×104(1.12) 0.284 (0.39) 0.022 (0.09)
Year no. 4 (dummy) 0.019 (0.135) 0.206 (0.98) 2.482 ×103(1.55) 0.146 (0.16) 0.157 (0.13)
Harvest no. 2 (dummy) 0.131 (0.337) 0.940 (2.79)∗∗∗ 4.009 ×103(2.28)∗∗ 1.084 (0.45) 1.204 (0.29)
Harvest no. 3 (dummy) 0.131 (0.337) 0.782 (2.31)∗∗ 2.484 ×103(1.34) 1.080 (0.45) 0.549 (0.26)
Harvest no. 4 (dummy) 0.126 (0.332) 0.563 (1.55) 1.116 ×103(0.49) 0.919 (0.42) 0.210 (0.15)
Harvest no. 5 (dummy) 0.119 (0.324) 0.640 (1.63) 4.019 ×103(1.78)0.372 (0.21) 0.690 (0.15)
Harvest no. 6 (dummy) 0.110 (0.313) 0.606 (1.45) 4.442 ×103(1.87)0.219 (0.13) 0.739 (0.10)
Harvest no. 7 (dummy) 0.100 (0.301) 0.433 (1.07) 3.510 ×103(1.45) 0.058 (0.04) 0.385 (0.16)
Harvest no. 8 (dummy) 0.086 (0.281) 0.188 (0.49) 2.458 ×103(1.11) 0.182 (0.17) 0.121 (0.13)
Harvest no. 9 (dummy) 0.054 (0.226) 0.217 (0.56) 2.472 ×103(1.14) 0.133 (0.12) 0.148 (0.07)
Harvest no. 10 (dummy) 0.012 (0.107) 0.359 (0.76) 5.636 ×103(2.31)∗∗ 0.346 (0.40) 0.785 (0.20)
Constant (ton/harvest-ha) 0.024 (0.38)
μ1.276 (6.40)∗∗∗
ρ0.956 (134.8)∗∗∗
Observations 429
Log likelihood 25.75
Pseudo R20.26
Notes: denotes significance at a 10% level, ∗∗ denotes significance at a 5% level, and ∗∗∗ denotes significance at a 1% level.
aThe marginal effects were evaluated for the mean values of the explanatory variables. Standard errors of the marginal effect statistics were calculated by bootstrap procedure. The effects of dummy variables are the difference
in the mean and variance under substitution of 0 and 1 in all observations at all bootstrap replications.
bDummies for years and fields are not shown.
by guest on May 22, 2014 from
744 April 2014 Amer. J. Agr. Econ.
Note that our estimation of the effect of
barn owls on alfalfa output is based on the
position of the occupied nesting boxes in
relation to the alfalfa fields. The box loca-
tions, however, might be non-random. That
is, Sde Eliyahu farmers may have intention-
ally located the boxes closer to fields with
higher alfalfa yields. To examine this possibil-
ity, we estimated the model again, this time
with the predation pressure in equation (15)
calculated as if all boxes are occupied (i.e.,
ξkt =1 for all k=1, ...,K). This resulted in
a lower value for the log-likelihood function
(LLF =12.38), and statistically insignificant
coefficients for the predation pressure and
its square. We therefore reject the hypothesis
that the positive impact of barn owls on the
output of alfalfa fields can be attributed to
the locations of the boxes rather than to the
locations of the occupied boxes only.
Returning to table 3, the effects of all other
explanatory variables that appear in both
Y(·)and A(·)on production and damage
abatement are opposite in sign. Thus, these
effects seem to offset each other such that
the marginal effects on mean output are all
statistically insignificant. With respect to the
variance of output, only the marginal effects
of irrigation and rainfall are statistically
significant, and both reduce yield volatility.
Flood irrigation has a direct negative effect
on alfalfa yield. This irrigation method is less
efficient than sprinkling, since much flood-
water is lost through deep-percolation flows.
However, flood irrigation affects abatement
positively. While we cannot reject the hypoth-
esis of zero marginal effect of irrigation on
mean alfalfa output, the coefficient of irri-
gation in Y(·)indicates a negative marginal
product. This can be explained by the irriga-
tion doses (see endnote 9), which, according
to the evaporation in the Beit She’an Valley
during the months of alfalfa growth (IMS
2013), are more than 25% higher than the
recommended doses (Tzukerman 2004).
Over-irrigation can result in reduced alfalfa
yield (Donovan and Meek 1983;Mueller,
Frate, and Campbell-Mathews 2007), and
is more likely to occur under conditions of
non-uniform infiltration (Feinerman, Letey,
and Vaux 1983), saline irrigation water that
requires excess irrigation for salt leaching
and low water prices, as is the case in the
Beit She’an Valley. As expected, organic
production tends to yield lower output than
conventional production. However, it also
stimulates abatement, possibly owing to
the avoidance of fertilization, which in turn
discourages weed growth. A longer period
between harvests increases output, but
decreases abatement, probably due to the
longer time afforded for the establishment
of populations of damaging agents. This may
also explain the lower abatement level in
the second year of production compared to
the first year (the effects are opposite in sign
to the coefficients). Productivity is higher
in the second and third harvests compared
to all other harvests. Larger fields are more
productive, but abatement is lower, indicating
a negative interaction between field size and
the barn owls’ effect.11 That is, as boxes are
usually located at the margins of the field,
the effect of barn owls on rodents in the inte-
rior areas of the field is lower, and further
decreases with increasing field size.
To test the sensitivity of our results to the
functional form specification adopted from
Saha, Shumway, and Havenner (1997),we
estimated a linear production function using
the same set of explanatory variables, in
which all variables were treated as produc-
tive inputs. The estimation results (online
appendix C), with respect to the signs and
significance levels of the coefficients of barn
owl predation pressure and its square, are
almost the same as those reported in table 3.
The two specifications yield almost iden-
tical estimates of the predation pressure’s
marginal effect. This finding reinforces
previous results by Carrasco-Tauber and
Moffitt (1992) and Lansink and Carpentier
(2001), which contradict Lichtenberg and
Zilberman’s (1986) hypothesis that damage-
abatement specifications lead to lower
estimates of the marginal effect of abate-
ment inputs. On the other hand, the linear
specification yields marginal effects with
higher significance levels for almost all of the
other explanatory variables.
With the estimated functions of predation
pressure, box selection and alfalfa produc-
tion, and additional price and cost data,12
11 The nature of the interaction between some two variables
Xand Qin the abatement process, calculated by 2G/∂XQ,
can be qualitatively represented by the sign of the product of
their coefficients, ωXωQ.
12 Barn owls affect revenues by changing per-hectare pro-
ductivity, and they entail fixed per-hectare costs associated with
by guest on May 22, 2014 from
Kan et al. Agricultural Rodent Control Using Barn Owls: Is it Profitable? 745
we are in a position to plug these elements
into equation (2) to evaluate the profitability
of biological rodent control by barn owls.
Four scenarios are compared. Scenario 1
represents the observed situation, in which
the 58 nesting boxes are in their current loca-
tions throughout the fields of Sde Eliyahu.
In scenario 2, we simulate alfalfa produc-
tion in the absence of all nesting boxes,
such that rodents are controlled only by the
aforementioned agronomic and natural fac-
tors. In scenario 3, we run an algorithm that
searches for the vector of optimal locations
of the 58 nesting boxes, x, which maximizes
the expected profits of the alfalfa fields.
The alfalfa revenues in this scenario are
calculated by:
(17) R=p1
where δi1stands for the probability of field
ibeing assigned to alfalfa production,
as computed based on our sample, and
E[y(Wi,Xi)]is the field’s output expecta-
tion as in equation (11). Scenario 4 is similar
to scenario 3, but instead of expected rev-
enues, the function to be maximized under
xincorporates the certainty-equivalent
CE =p1
where V[y(Wi,Xi)]is the output’s variance
expressed by equation (12), and is the
Arrow-Pratt measure of relative risk aver-
sion, which was evaluated by Bar-Shira, Just,
and Zilberman (1997) to be 0.611 for farm-
ers in Israel. As barn owls increase both the
production expectation and the variance
the installation and maintenance of nesting boxes. The output
price is $264/ton, as reported by the Israel Field Crops Growers
Association (2010), for alfalfa under conventional production.
Variable costs associated with harvesting and hauling amount to
$38/ton (IMARD 2010). The per-box costs were estimated at
$50/year, based on an installation cost of $250, a 10-year lifetime,
with one renovation at a cost of $60 and 0.1 working days per
year for monitoring and cleaning. Attributing the costs of all 58
nesting boxes to the 42 ha allocated to alfalfa in Sde Eliyahu in
an average year, we get a cost of $69/ha per year.
(see table 3), the certainty-equivalent rev-
enue captures their counter effects on the
production and risk premium.
The objective of scenarios 3 and 4 is to get
an idea of how much higher the profits of
Sde Eliyahu’s alfalfa fields could be if the 58
boxes were originally located so as to maxi-
mize those profits. These scenarios, however,
are associated with the challenge of solving a
complicated non-linear spatial optimization
problem, and are based on extrapolations of
our estimated functions. Therefore, simplifi-
cations and constraints are needed to obtain
practical and computable results.
Our estimation of the alfalfa production
function Hi1(ri1(u,x,δ,a,h),vi,b1)implies
that alfalfa outputs would considerably
increase with cumulative predation pressure
(table 3). The estimated predation-pressure
function likj(dik (ui,xk)) tells us that the cumu-
lative pressure on alfalfa fields will increase
convexly as the distance between those fields
and occupied nesting boxes decreases. The
proximity of nesting boxes to alfalfa fields
may also increase their occupancy rate, as
can be learned from the estimated nesting
selection function sk(u,x,δ,ak,hk)in table 2.
Integrating these three effects implies that
profits would be maximized if as many nest-
ing boxes as possible were to be located as
close as feasible to the alfalfa fields. A coun-
teracting factor is the occupancy rate, which
decreases when boxes become too close
to each other. Nesting rates may also be
restricted by the impact of the boxes’ dis-
tances from other land uses, such as date
trees. However, while our model captures
these opposing forces, the reliability of our
predictions is expected to diminish as we
extrapolate further. For instance, nesting
rates may be limited by unobserved vari-
ables such as overall barn owl population in
the relevant area, or competition with other
predators such as jackals, kestrels, and wild-
cats, which may affect the barn owls’ hunting
success. In addition, the location of the boxes
should account for operational farming prac-
tices, such as the movement of cultivation
machinery. These considerations can be taken
into account in the model by introducing
constraints. The two optimization scenar-
ios incorporate two additional constraints:
(a) all of the boxes are restricted to being
located at the borders of the fields, at least
100 m apart, and (b) the per-hectare alfalfa
production in every field is restricted to no
more than 25 ton/year, which is 25% higher
by guest on May 22, 2014 from
746 April 2014 Amer. J. Agr. Econ.
Figure 1. Optimal (scenarios 3 and 4) versus current (scenario 1) distribution of nesting
boxes in relation to alfalfa crop-rotation fields
than the typical alfalfa productivity reported
by IMARD (2010). In addition, to facilitate
computation, our optimization algorithm
considers 1,500 predefined potential points
(located at the borders of all of Sde Eliyahu’s
fields, 100 m apart), and searches among them
for the optimal location of one box at a time.
This is a stepwise procedure in which the first
box is located at the optimal point given that
it is the only one in the area; the second box
is optimally located given the location of the
first box, and so on.13
Apparently, scenarios 3 and 4 both yield a
similar solution for x. Figure 1shows current
versus optimal nesting box distributions rel-
ative to the fields with positive probabilities
of being assigned to alfalfa throughout crop
The simulation results are summarized in
table 4.
Scenario 2 constitutes a benchmark for
the calculation of the contribution of barn
13 Finding the optimal solution would require computing the
value of the objective functions for each of the 2.3 ×10105[=
1500!/(581442!)]combinations of 58 boxes placed in 1,500 loca-
tions. While our algorithm applies only 85,347 [= (1501 ×1500
1443 ×1442)/2]computations, and therefore may not hit upon
the optimal location, it is computationally feasible, and believed
to be satisfactory for evaluating the extent to which profits of
alfalfa fields could be increased.
owls’ rodent control to production and profit
under the other three scenarios. The yield
expectation under scenario 2 is computed by
the use of equation (11), while substituting
ri1=0 for all i=1, ...,Iinto equation (15)
and holding all other variables at their time-
average levels. The alfalfa output attributable
to the presence of the 58 nesting boxes in
their current locations equals the difference
between the expected productions under
scenarios 1 and 2. This calculation results in
a contribution of 1.35 ton/ha per year, which
constitutes 9.4% of the observed production
expectation of 14.38 ton/ha per year. The
associated profit contribution amounts to
$235.8/ha per year. The calculated profit con-
tribution based on the certainty-equivalent
profits is slightly higher ($245.0/ha per year);
this is because the increase in the production
expectation exceeds that of the variance.
Thus, rodent control by barn owls is found
to be profitable.14 As noted, these are likely
14 The profitability of alfalfa production is rather small, and
may even be negative in certain years, so that an increase of 9.4%
in production can make a significant difference in terms of profits.
Based on production studies published by extension specialists at
UC Davis (2008),Texas AgriLife Extension Service (2011), and
the University of Wisconsin (2011), an output increase of 9.4%
implies profit increases of 31%, 67%, and 107%, respectively. A
similar study published by Iowa State University (2011) found
by guest on May 22, 2014 from
Kan et al. Agricultural Rodent Control Using Barn Owls: Is it Profitable? 747
Table 4. Rodent-Control Scenarios
Scenario 1 Scenario 2 Scenarios3&4
The 58 Nesting The 58 The 58 Nesting Boxes Are
Boxes Are in Nesting Located So as to Maximize
Their Current Boxes Are Alfalfa Expected Profits and
Description Locations Eliminated Certainty-Equivalent Profits
Average distance between
box and all other boxes
1.00 – 1.61
Average occupancy rate of
0.46 – 0.61
Average pressure on alfalfa
fields (Probability/ha)
4.54 ×1050.00 1.35 ×104
Average alfalfa production
expectation (ton/ha per
14.38 13.03 17.07
Average standard
deviation of production
(ton/ha per year)
8.11 7.63 9.29
Average certainty
equivalent production
(ton/ha per year)
13.17 11.78 15.68
Profit increase compared
to Scenario 2 ($/ha per
Expected profits 235.8 0.00 845.2
245.0 0.00 812.9
to be underestimates of the contribution of
barn owls to overall profitability, since we
completely ignore the potential contribution
of the nesting boxes to the yields of other
crops, most of which are more attractive for
barn owl hunting than alfalfa (see table 1).
Compared with the current box locations
(scenario 1), boxes in scenarios 3 and 4 are
located around the fields with high proba-
bilities of being assigned to alfalfa (figure 1),
and the average nesting rate is consider-
ably higher than the observed nesting rate
(table 4). Consequently, the average per-
hectare predation pressure on alfalfa fields
is an order of magnitude higher; thus, the
portion of the production associated with
the presence of barn owls increases from
9.4% to more than 23%, and the computed
contribution of the barn owls to alfalfa prof-
its is 3.3 times of that under the observed
situation. These results highlight the con-
siderable impact of the locations of nesting
net losses in alfalfa production, yet an output increase of 9.4%
could have reduced the losses by 22%. Similarly,a study provided
by IMARD (2010) also found net losses in alfalfa production
in Israel, and here a 9.4% yield increase could reduce losses by
boxes on alfalfa profits. In a broader per-
spective, our case study of rodent control
using barn owls illustrates the importance of
the spatial distribution of sources of agro-
biological agents (e.g., honeybee apiaries) as
a farming-management tool.
The question arises of how robust these
evaluations are to functional form specifica-
tions. We examine this question with respect
to the production function by simulating
scenarios 1, 2, and 3 using the aforemen-
tioned linear specification. Apparently, unlike
the marginal effect, the linear production
function yields evaluations of barn owls’
contributions to outputs and profits that are
higher under both the observed and optimal
locations of the boxes (online appendix C).
This finding indicates that the evaluation
obtained using the functional form of damage
abatement is rather conservative.
As indicated by scenarios 3 and 4, if alfalfa
is assumed to be the only crop whose profit
can be increased by barn owl activity, the
current spatial distribution of nesting boxes
at Sde Eliyahu is not optimal; the returns
on some of the boxes may not even cover
their installation and maintenance costs.
To examine this issue further, we applied
by guest on May 22, 2014 from
748 April 2014 Amer. J. Agr. Econ.
0 102030405060
$/box/ha per year
Current box locations
Optimal box locations
Figure 2. Marginal profits of the number of
boxes under current and optimized locations
our stepwise optimization algorithm to the
58 observed locations of the boxes. This
enabled us to compute the marginal profit
of each additional nesting box, as presented
in figure 2. A similar curve is presented for
the optimal box locations, as selected by the
optimization algorithm under scenarios 3
and 4.15 As suspected, in their current loca-
tions, 10 of the nesting boxes do not cover
their costs. Nevertheless, as already noted,
the array of boxes as a whole is still prof-
itable. While the marginal profit curve under
scenarios 3 and 4 fluctuates noticeably and
exhibits a decreasing trend, all of the boxes
are profitable.
Our empirical application has two policy
implications. First, the results indicate that
rodent control by barn owls is profitable
from a farming point of view. Thus, under a
hypothetical case in which rodenticides are
absolutely prohibited, aside from informing
farmers about the potential profitability of
the method and developing guidance and
training programs, additional governmen-
tal intervention to promote adoption of the
method (e.g., by policy instruments such as
15 The curves exhibit non-monotonic patterns due to the spatial
interrelations among the boxes; each additional box can affect
the occupancy probability of the boxes located earlier by the
stepwise optimization algorithm, and thus yield a larger marginal
profitability compared to the previous box.
subsidies) is unnecessary.16 However, more
active governmental intervention may be
warranted if rodenticides are allowed, or in
areas where farmers own small agricultural
plots where spillover effects of the barn owl
damage-abatement services may lead to
free riding and thus to the placement of a
suboptimal number of nesting boxes. Sec-
ond, despite the fact that we cannot directly
compare the profitability of rodent control
by barn owls to that of rodent control by
rodenticides, our analysis evaluates profit
contributions that are both significant and
can be considered underestimates of the
overall returns stemming from barn owl dam-
age control. Thus, if policy-makers such as
those in Israel (IMARD 2011) are looking to
reduce the considerable environmental dam-
age caused by rodenticides (e.g., Yom-Tov
and Mendelssohn 1988;Zurita et al. 2007),
our findings provide strong arguments for
more severely restricting the regulatory
conditions under which rodenticides are
This study leaves a good deal of room for
future research. For example, data on the
various ecological, zoological, and economical
components of the barn owl system can be
collected at a finer resolution to elucidate
the costs and benefits of control by barn owls
versus alternative actions. The barn owls’
contribution to profit may be assessed with
respect to more crops, which would enable
the computation of a more realistic opti-
mal spatial distribution of boxes, including
a determination of the optimal number of
boxes. The profitability of the method should
also be compared to that of rodenticides.
Finally, our application evaluates contri-
butions to farmers’ profits only; designing
rodent-control policies based on a wider
social perspective would require valuations
of the environmental damage abated through
the avoidance of rodenticide use, the benefits
associated with preserving barn owls, and the
impacts of barn owls on other endangered
The spatial economic model developed
in this study is applicable to other agrobio-
logical systems, particularly those associated
with point-source spatial impacts. Examples
16 In recent years, the method has been rapidly adopted by
Israeli farmers, partly owing to the governmental training activ-
ities and financial support for the installation, maintenance, and
monitoring of nesting boxes (IMEP 2009;Motro et al. 2010).
by guest on May 22, 2014 from
Kan et al. Agricultural Rodent Control Using Barn Owls: Is it Profitable? 749
include the selection of locations for hon-
eybee apiaries, which influence pollination
services and honey production (Manning
and Wallis 2005), the management of patches
of non-crop habitats to enhance natural
pest control (Bianchi, Booij, and Tscharntke
2006), and the positions of cattle watering
points, which affect the spatial variation of
vegetation in rangelands and thus meat pro-
duction (Ludwig et al. 1999). Nevertheless,
extensions are required to further expand
the applicability of the model, particularly to
cases in which various movement processes,
such as dispersal and wandering, are impor-
tant (Nathan et al. 2008), for example in the
release of lady beetles (Baker et al. 2003) and
sterile flies (Enkerlin 2007). Introducing the
time dimension might also enable capturing
external stochastic spatiotemporal effects
(Harper and Zilberman 1989), the dynamics
of predator-prey systems (Rafikov, Balthazar,
and von Bremen 2008), and long-term adap-
tation and resistance development in pests
(Hueth and Regev 1974).
Supplementary material
Supplementary material is available at
Askham, L.R. 1990. Effect of Artificial
Perches and Nests in Attracting Rap-
tors to Orchards. Proceeding of the
Fourteenth Vertebrate Pest Conference.
Babcock, B.A., E. Lichtenberg, and D. Zilber-
man. 1992. Impact of Damage Control
and Quality of Output: Estimating Pest
Control Effectiveness. American Jour-
nal of Agricultural Economics 74 (1):
Baker, S.C., J.A. Elek, R. Bashford, S.C.
Paterson, J. Madden, and M. Battaglia.
2003. Inundative Release of Coccinel-
lid Beetles into Eucalypt Plantations
for Biological Control of Chrysomelid
Leaf Beetles. Agricultural and Forest
Entomology 5: 97–106.
Bar-Shira, Z., R.E. Just, and D. Zilberman.
1997. Estimation of Farmers’ Risk
Attitude: An Econometric Approach.
Agricultural Economics 17 (2–3):
Bianchi, F.J.J.A., C.J.H. Booij, and T. Tscharn-
tke. 2006. Sustainable Pest Regulation
in Agricultural Landscapes: A Review
on Landscape Composition, Biodiversity
and Natural Pest Control. Proceedings
Biological Sciences 273: 1715–1727.
Blackwell, M., and A. Pagoulatos. 1992. The
Econometrics of Damage Control. Amer-
ican Journal of Agricultural Economics
74 (4): 1040–1044.
Brown, C., L. Lynch, and D. Zilberman. 2002.
The Economics of Controlling Insect-
transmitted Plant Diseases. American
Journal of Agricultural Economics 84 (2):
Burnham, K.P. and D.R. Anderson. 2002.
Model Selection and Multimodel Infer-
ence: A Practical Information-Theoretic
Approach. Second Edition. New York:
Carrasco-Tauber, C. and L.J. Moffitt. 1992.
Damage Control Econometrics: Func-
tional Specification and Pesticide
Productivity. American Journal of
Agricultural Economics 74 (1): 158–162.
Chambers, R.G., and E. Lichtenberg.
1994. Simple Econometrics of Pesti-
cide Productivity. American Journal of
Agricultural Economics 76 (3): 407–417.
Charter, M., I. Izhaki, K. Meyrom, Y. Motro,
and Y. Leshem. 2009. Diets of Barn
Owls Differ in the Same Agricultural
Region. Wilson Journal of Ornithology
121: 378–383.
Charter, M., K. Meyrom, Y. Leshem, S. Aviel,
I. Izhaki, and Y. Motro. 2010. Does Nest
Box Location and Orientation Affect
Occupation Rate and Breeding Success
of Barn Owls Tyto alba in a Semi-Arid
Environment? Acta Ornithologica 45 (1):
Charter, M., Y. Leshem, K. Meyrom, O. Peleg,
and A. Roulin. 2012. The Importance
of Micro-Habitat in the Breeding of
Barn Owls Tyto alba.Bird Study 59 (3):
Cox, P.R., and R.H. Smith. 1990. Rodenticide
Ecotoxicology: Assessing Non-target
Population Effects. Functional Ecology 4
(3): 315–320.
Davis, S.A., H. Leirs, R. Pech, Z. Zhang, and
N.C. Stenseth. 2004. On the Economic
Benefit of Predicting Rodent Outbreaks
in Agricultural Systems. Crop Protection
23: 305–314.
Donovan, T.J., and B.D. Meek. 1983. Alfalfa
Responses to Irrigation Treatment and
by guest on May 22, 2014 from
750 April 2014 Amer. J. Agr. Econ.
Environment. Agronomy Journal 75 (3):
Duckett, J.E. 1976. Owls as Major Predators
of Rats in Oil Palm Estates with Partic-
ular Reference to the Barn Owl (Tyto
alba). Planter Kuala Lumpur 52: 4–15.
Eddleston, M. 2000. Patterns and Prob-
lems of Deliberate Self-poisoning in the
Developed World. Quarterly Journal of
Medicine 93: 715–731.
Enkerlin, W. 2007. Guidance for Packing,
Shipping, Holding and Release of Sterile
Flies in Area-wide Fruit Fly Control Pro-
grams. Plant production and protection
paper 190. Rome: FAO Publications.
Feinerman, E., J. Letey, and H.J. Vaux, Jr.
1983. The Economics of Irrigation with
Nonuniform Infiltration. Water Resources
Research 19 (6): 1410–1414.
Fox, G., and A. Weersink. 1995. Damage Con-
trol and Increasing Returns. American
Journal of Agricultural Economics 77 (1):
Frey, C., C. Sonnay, A. Dreiss, and A. Roulin.
2011. Habitat, Breeding Performance,
Diet and Individual Age in Swiss Barn
Owls (Tyto alba). Journal of Ornithology
152: 279–290.
Griffiths, G.J.K., J.M. Holland, A. Bailey,
and M.B. Thomas. 2008. Efficacy and
Economics of Shelter Habitats for Con-
servation Biological Control. Biological
Control 45 (2): 200–209.
Harper, C.R., and D. Zilberman. 1989.
Pest Externalities from Agricultural
Inputs. American Journal of Agricultural
Economics 71: 692–702.
Hueth, D., and U. Regev. 1974. Optimal
Agricultural Pest Management with
Increasing Pest Resistance. American
Journal of Agricultural Economics 56:
Iowa State University. 2011. Alfalfa or
Alfalfa-grass Hay Budget. http://www.
Israel Ministry of Agriculture and Rural
Development (IMARD). 2010.
Production-Instruction for Growers
(Hebrew). http://www.shaham.moag.
———. 2011. Green Agriculture (Hebrew).
Israel Ministry of Environmental Protection
(IMEP). 2009. Barn Owls and Kestrels:
Biological Pest Control Agents. Israel
Environmental Bulletin 35: 27–29.
Israel Field Crops Growers Associa-
tion. 2010. Prices for Hay (Hebrew).
Israeli Meteorological Service (IMS).
2013. Monthly Average Pan Evapora-
tion (Hebrew).
Kay, B., L. Twigg,T. Korn, and H. Nicol. 1994.
The Use of Artificial Perches to Increase
Predation on House Mice by Raptors.
Wildlife Research 21: 95–106.
Lansink, A.O, and A. Carpentier. 2001. Dam-
age Control Productivity: An Input
Damage Abatement Approach. Journal
of Agricultural Economics 52 (3): 11–22.
Leirs, H. 2003. Management of Rodents in
Crops: The Pied Piper and His Orchestra.
In Rats, Mice, and People: Rodent Biology
and Management, ed. Singleton, G.R.,
L.A. Hinds, C.J. Krebs, and D.M. Spratt,
183–190. Canberra: ACIAR.
Leshem, Y., D. Alon, I. Atrash, A. Marei,
M. Abu Rashid, I. Kan, and I. Izhaki.
2010. Agricultural Rodent Pest Control
by Avian Predators: An Ecologically
Sustainable Approach. Research report,
USAID, TA-MOU-06-M25-078. http://
Lessells, C.M., and D.W. Stephens. 1983. Cen-
tral Place Foraging: Single-Prey Loaders
Again. Animal Behavior 31: 238–243.
Lichtenberg, E., and D. Zilberman. 1986.
The Econometrics of Damage Control:
Why Specification Matters. American
Journal of Agricultural Economics 68 (2):
Ludwig, J.A., R.W. Eager, R.J. Williams,
and L.M. Lowe. 1999. Declines in
Vegetation Patches, Plant Diversity,
and Grasshopper Diversity Near Cat-
tle Watering-points in the Victoria
River District, Northern Australia. The
Rangeland Journal 21 (1): 135–149.
Manning, R., and I.R. Wallis. 2005. Seed
Yields in Canola (Brassica napus
cv. Karoo) Depend on the Distance
of Plants from Honeybee Apiaries.
Australian Journal of Experimental
Agriculture 45 (10): 1307–1313.
Moran, S., and H. Keidar. 1993. Checklist
of Vertebrate Damage to Agricul-
ture in Israel. Crop Protection 12:
by guest on May 22, 2014 from
Kan et al. Agricultural Rodent Control Using Barn Owls: Is it Profitable? 751
Motro, Y., Y. Leshem, S. Aviel, M. Charter,
D. Alon, and Y. Chassin. 2010. The Use
of Barn Owls and Kestrels as Biological
Rodent Controllers in Agriculture. Soci-
ety for the Protection of Nature, Ministry
of Environmental Protection, Ministry
of Agriculture and Rural Development,
Israel (Hebrew).
Mueller, S.C., C.A. Frate, and M. Campbell-
Mathews. 2007. Alfalfa Stand Establish-
ment. In Irrigated Alfalfa Management
for Mediterranean and Desert Zones,
ed. Summers, C.G., and D.H. Putnam,
Chapter 4. UC ANR Publication 8290.
Nathan, R., W.M. Getz, E. Revilla, M.
Holyoak, R. Kadmon, D. Saltz, and
P.E. Smouse. 2008. A Movement Ecol-
ogy Paradigm for Unifying Organismal
Movement Research. Proceedings of
the National Academy of Sciences of
the United States of America 105 (49):
Nordblom, T.L., M.J. Smyth, A. Swirepik,
A.W. Sheppard, and D.T. Briese. 2002.
Spatial Economics of Biological Control:
Investing in New Releases of Insects for
Earlier Limitation of Paterson’s Curse
in Australia. Agricultural Economics 27:
Parshad, V.R. 1999. Rodent Control in India.
Integrated Pest Management Reviews 4:
Polasky, S., E. Nelson, E. Lonsdorf, P. Fack-
ler, and A. Starfield. 2005. Conserving
Species in a Working Landscape: Land
Use with Biological and Economic
Objectives. Ecological Applications 15
(4): 1387–1401.
Polasky, S., E. Nelson, D. Pennington, and
K.A. Johanson. 2011. The Impact of
Land-use Change on Ecosystem Services,
Biodiversity and Returns to Landowners:
A Case Study in the State of Minnesota.
Environmental and Resource Economics
48: 219–242.
Proulx, G. 1998. Evaluation of Strychnine
and Zinc Phosphide Baits to Control
Northern Pocket Gophers (Thomomys
talpoides) in Alfalfa Fields in Alberta,
Canada. Crop Protection 17: 135–138.
Rafikov, M., J.M. Balthazar, and H.F. von
Bremen. 2008. Mathematical Model-
ing and Control of Population Systems:
Applications in Biological Pest Control.
Applied Mathematics and Computation
200: 557–573.
Saha, A., C.R. Shumway, and A. Havenner.
1997. The Economics and Econometrics
of Damage Control. American Journal of
Agricultural Economics 79 (3): 773–785.
Singleton, G.R. 2003. Impacts of Rodents
on Rice Production in Asia. IRRI
Discussion Paper Series No. 45, Los
Banos, Philippines.
Skonhoft, A., H. Leirs, H.P. Andreassen, L.A.
Mulungu, and N.C. Stenseth. 2006. The
Bioeconomics of Controlling an African
Rodent Pest Species. Environment and
Development Economics 11: 453–475.
Stenseth, N.C., H. Leirs, A. Skonhoft, S.A.
Davis, R.P. Pech, H.P. Andreassen, G.R.
Singleton, et al. 2003. Mice, Rats, and
People: The Bio-economics of Agricul-
tural Rodent Pests. Frontiers in Ecology
and the Environment 1 (7): 367–375.
Taylor, I. 1994. Barn Owls: Predator-
prey Relationships and Conservation.
Cambridge: University Press.
Texas AgriLife Extension Service. 2011.
Alfalfa Production, Irrigated. http://age
Thirgood, S.J., S.M. Redpath, and I.M.
Graham. 2003.What Determines the
Foraging Distribution of Raptors on
Heather Moorland? OIKOS 100: 15–24.
Tzukerman, E. 2004. Recommendations
for Sawing and Growing Alfalfa for
Animal Feed (Hebrew). http://www.
University of California, Davis. 2008. Sam-
ple Costs to Establish and Produce
Alfalfa Hay. http://costStudies.ucDavis.
University of Wisconsin Center for Dairy
Profitability. 2011. Established Alfalfa
Harvested as Dry Hay. http://cdp.wisc.
edu/Budgets/2011/120/Established Alfalfa
Harvested as Haylage.pdf
Venables, W.N., and B.D. Ripley. 2002. Mod-
ern Applied Statistics with S. Fourth
Edition. New York: Springer.
Wood, B.J., and C.G. Fee. 2003. A Critical
Review of the Development of Rat Con-
trol in Malaysian Agriculture Since the
1960s. Crop Protection 22: 445–461.
Wooldridge, J.M. 2005. Simple Solutions
to the Initial Conditions Problem in
Dynamic, Nonlinear Panel Data Models
with Unobserved Heterogeneity. Journal
of Applied Econometrics 20: 39–54.
by guest on May 22, 2014 from
752 April 2014 Amer. J. Agr. Econ.
Yom-Tov, Y. and H. Mendelssohn. 1988.
Changes in the Distribution and Abun-
dance of Vertebrates in Israel During the
20th Century. In The Zoogeography of
Israel, ed. Yom-Tov, Y. and E. Tchernov,
514–547. Dordrecht, Holland: Dr. W.
Zhang, W., T.H. Ricketts, C. Kremen, K.
Carney, and S.M. Swinton. 2007. Ecosys-
tem Services and Dis-services to
Agriculture. Ecological Economics 64
(2): 253–260.
Zurita, J.L., A. Jos, A.M. Camea’n, M.
Salguero, M. Lo’pez-Artι’guez, and
G. Repetto. 2007. Ecotoxicological Eval-
uation of Sodium Fluoroacetate on
Aquatic Organisms and Investigation
of the Effects on Two Fish Cell Lines.
Chemosphere 67:1–12.
by guest on May 22, 2014 from
... Labuschagne et al. (2016) carried out a literature review to assess the effectiveness of biological control by raptors for rodent pest management in agriculture, and found that most of the literature focused on barn owls but also that there are too few studies, and of those the minority showed scientific rigor. Many studies reported the food habits of barn owls in agricultural landscapes (e.g., Charter et al., 2009;Kross et al., 2016) but few reported actual impact on pest populations, crop damage and economic implications (but see Motro, 2011;Kan et al., 2013;Browning et al., 2016). Kan et al. (2013) found that in Beit-She'an Valley, a semi-arid agricultural environment in the north of Israel, alfalfa yields could increase from 13.03 (if no nesting boxes were in place) to 17.07 tons per hectare if the existing 58 barn owl nesting boxes would be placed optimally (compared to 14.38 tons per hectare under the current locations of the nesting boxes). ...
... Many studies reported the food habits of barn owls in agricultural landscapes (e.g., Charter et al., 2009;Kross et al., 2016) but few reported actual impact on pest populations, crop damage and economic implications (but see Motro, 2011;Kan et al., 2013;Browning et al., 2016). Kan et al. (2013) found that in Beit-She'an Valley, a semi-arid agricultural environment in the north of Israel, alfalfa yields could increase from 13.03 (if no nesting boxes were in place) to 17.07 tons per hectare if the existing 58 barn owl nesting boxes would be placed optimally (compared to 14.38 tons per hectare under the current locations of the nesting boxes). ...
Rodent agricultural pests cause significant food loss every year. Attempts at mitigation via chemical pest control may cause secondary poisoning and harm non‐target species. Biological pest control by bolstering barn owl Tyto alba populations through the provision of artificial nest boxes is in use in several countries. The national biological pest control project of Israel began in the early 1980s in the Mediterranean zone and was subsequently expanded to the northwestern Negev desert, including areas adjacent to nature reserves and natural sand dunes, a threatened habitat in Israel. We analyzed prey of barn owls in the northwestern Negev to determine whether owls preyed on non‐target endemic, threatened rodents. A total of 14 632 prey items were collected from 95 nesting boxes between 2013 and 2016. We found that barn owls feed on protected and locally endangered species such as Gerbillus andersoni allenbyi (vulnerable − VU), Gerbillus pyramidum (VU), the endemic Meriones sacramenti (endangered) and Gerbillus gerbillus (critically endangered). These species constitute a significant proportion of barn owl diets (sometimes more than half of the prey items in a single nest box), especially in areas under 5 km from sand dunes, suggesting that bolstering the barn owl population may threaten locally endangered species. It could be hypothesized that agricultural crops serve as a resource for endangered species, allowing their numbers to increase, in which case their conservation status should be reassessed. However, extensive rodent trappings in agricultural fields in the region revealed <1% Gerbillus individuals, while in nearby dunes only gerbils were trapped. Insufficient data are available to assess whether or not M. sacramenti could have become a local pest. Our study highlights the risks that may be associated with the use of barn owls as rodent control agents in areas where natural open landscapes and nature reserves are interspersed with agricultural landscapes.
... The need to control rodent pests has led to an interest in integrated pest management using biological control agents, primarily barn owls (Labuschagne et al. 2016). By installing nest boxes in vineyards, farmers can attract barn owls, which may be able to act as natural predators and reduce populations of rodent pests (Meyrom et al. 2009, Paz et al. 2012, Kan et al. 2013, Browning et al. 2017. However, barn owls are mobile predators and they show a preference for open natural habitats (Taylor 1994), so the effectiveness of nest box installation on pest control delivery may depend on the composition of landscapes surrounding wine grape vineyards (Chart er et al. 2012). ...
Full-text available
Abstract Landscape composition can strongly affect the delivery of ecosystem services in agroecosystems. Conserving uncultivated habitats can support ecosystem services, but in Mediterranean biomes, these lands can also increase the area susceptible to wildfires. In the world‐renowned wine‐producing region of Napa Valley, California, wine grape growers install nest boxes to attract American barn owls (Tyto furcata), which may reduce rodent crop damage. Annual monitoring of 273 nest boxes began in 2015, and devastating wildfires burned approximately 60,000 ha in the region in 2017, including homes and businesses, as well as some vineyards and uncultivated land. The goal of this study was to determine whether changes in nest box occupancy were attributed to wildfires, nest box design, land cover type, or some combination of these variables. Occupancy surveys before and after these wildfires revealed changes in habitat selection at the nest scale. Occupancy increased during the study, reaching its highest point after the fires. Owls were found breeding in recently burned areas that were previously unoccupied and modeling results showed that nest box occupancy had a positive relationship with burned areas, particularly with edges of the fire perimeter. Barn owls also consistently showed a strong preference for taller nest boxes that were surrounded by more grassland than other land cover types and a moderate selection for wooden over plastic boxes. These results illustrate an incentive for the conservation of uncultivated habitat, particularly grassland, in vineyard ecosystems, and they provide an example of a mobile pest predator’s response to wildfire disturbance. In this case, results suggest an agroecosystem service made resilient to wildfire by the owls’ selection of burned and uncultivated habitats.
... Another experiment in the same settings revealed high variation in capture duration (from first attack to a successful capture) of spiny mice (Acomys cahirinus) and Günther's voles (Microtus guentheri), ranging from 0.5 sec to 43 min (63). The longest prey captures occurred while hunting voles, the preferred prey of owls in Israel (65,66) and in the Hula Valley in particular (67). Thus, as it is hard to catch a moving target, adopting irreproducible (and thus unpredictable) movement tactics may prove beneficial for a predator. ...
Movement tracks of wild animals frequently fit models of anomalous rather than simple diffusion, mostly reported as ergodic superdiffusive motion combining area-restricted search within a local patch and larger-scale commuting between patches, as highlighted by the L\'evy walk paradigm. Since L\'evy walks are scale invariant, superdiffusive motion is also expected within patches, yet investigation of such local movements has been precluded by the lack of accurate high-resolution data at this scale. Here, using rich high-resolution movement datasets ($>\! 7 \times 10^7$ localizations) from 70 individuals and continuous-time random walk modeling, we found subdiffusive behavior and ergodicity breaking in the localized movement of three species of avian predators. Small-scale, within-patch movement was qualitatively different, not inferrable and separated from large-scale inter-patch movement via a clear phase transition. Local search is characterized by long power-law-distributed waiting times with diverging mean, giving rise to ergodicity breaking in the form of considerable variability uniquely observed at this scale. This implies that wild animal movement is scale specific rather than scale free, with no typical waiting time at the local scale. Placing these findings in the context of the static-ambush to mobile-cruise foraging continuum, we verify predictions based on the hunting behavior of the study species and the constraints imposed by their prey.
... There is a wealth of literature on pesticide use and its determinants (Fan et al., 2020, Liu and Huang, 2013, Möhring et al., 2020b, Möhring et al., 2020c, Serra et al., 2005, Sunding and Zivin, 2000. Earlier literature has also addressed pest prevention measures, their adoption and interdependencies with pesticide use (Brown, 2018, Kan et al., 2013, Olson and Roy, 2005. However, the link between extension source, adoption of preventive measures and pesticide applications has not been empirically established in the literature so far. ...
Full-text available
Does it matter whether farmers receive advice on pest management strategies from public or from private (pesticide company affiliated) extension services? We use survey data from 733 Swiss fruit growers who are currently contending with an infestation by an invasive pest, the fruit fly Drosophila Suzukii. We find that farmers who are advised by public extension services are more likely (+9-10%) to use preventive measures (e.g. nets) while farmers who are advised by private extension services are more likely (+8-9%) to use synthetic insecticides. These results are robust to the inclusion of various covariates, ways to cluster standard errors, and inverse probability weighting. We also show that our results are unlikely to be driven by omitted variable bias. Our findings have implications for the current debates on both the ongoing privatization of agricultural extension and concerns regarding negative environmental and health externalities of pesticide use.
... The study area reported nine raptors, including T. alba. This raptor has an extensive global distribution and is a well-known predator of rodents, providing an economic service to farmers by controlling the rodent population in farmlands (Marti et al., 2005;Kan et al., 2013;Kross et al., 2016). Tyto alba is nocturnal (Taylor, 1994;Abramsky et al., 1996) and can hunt nocturnal rodents in the farmlands within the study area. ...
Full-text available
In this paper, we consider a fractional-order eco-epidemic model based on the Rosenzweig– MacArthur predator–prey model. The model is derived by assuming that the prey may be infected by a disease. In order to take the memory effect into account, we apply two fractional differential operators, namely the Caputo fractional derivative (operator with power-law kernel) and the Atangana–Baleanu fractional derivative in the Caputo (ABC) sense (operator with Mittag–Leffler kernel). We take the same order of the fractional derivative in all equations for both senses to maintain the symmetry aspect. The existence and uniqueness of solutions of both eco-epidemic models (i.e., in the Caputo sense and in ABC sense) are established. Both models have the same equilibrium points, namely the trivial (origin) equilibrium point, the extinction of infected prey and predator point, the infected prey free point, the predator-free point and the co-existence point. For a model in the Caputo sense, we also show the non-negativity and boundedness of solution, perform the local and global stability analysis and establish the conditions for the existence of Hopf bifurcation. It is found that the trivial equilibrium point is a saddle point while other equilibrium points are conditionally asymptotically stable. The numerical simulations show that the solutions of the model in the Caputo sense strongly agree with analytical results. Furthermore, it is indicated numerically that the model in the ABC sense has quite similar dynamics as the model in the Caputo sense. The essential difference between the two models is the convergence rate to reach the stable equilibrium point. When a Hopf bifurcation occurs, the bifurcation points and the diameter of the limit cycles of both models are different. Moreover, we also observe a bistability phenomenon which disappears via Hopf bifurcation.
Large-scale conversion of uncultivated land to agriculture threatens wildlife and can diminish ecosystem services provided by nature. Understanding how wildlife provision ecosystem services may incentivize wildlife conservation in agricultural landscapes. Attracting Barn Owls (Tyto furcata) to nest on farms for pest management has been implemented worldwide but has not been evaluated in vineyard agroecosystems. Napa Valley, California, is a renowned winegrape growing region, and viticulturists encourage Barn Owl occupancy to help minimize damage from Botta’s pocket gophers (Thomomys bottae) and voles (Microtus spp.). This study modeled the use of habitats in space and time by hunting Barn Owls, providing information about their potential to provide the critical ecosystem service of pest consumption. We used global positioning system tags to track hunting owls and used a resource selection function to compare used and available habitats. We constructed the intensity of use and home range-movement maps using a time local convex hull analysis from location data. We found that Barn Owls selected uncultivated habitats when hunting, some of which were relatively rare on the landscape. Approximately, one-third of Barn Owl hunting locations occurred in vineyards, but this use was out of proportion to the availability of vineyards, which comprised 50% of the area around nest boxes. The owls’ use of vineyards increased with decreasing amount of selected uncultivated habitat in the landscape. However, as reported by a previous study, the occupancy of nest boxes in vineyards increases with uncultivated habitats nearby. Future research should model landscape composition to determine the amount of preferred habitat necessary to support occupancy as well as hunting in vineyards. A true test of pest management by Barn Owls awaits experimentation coupled with monitoring rodent populations.
We analysed the diet of the Burrowing Owl (Athene cunicularia) and the American Kestrel (Falco sparverius) in a fragmented habitat and natural surroundings in the arid ecosystem of the Baja California peninsula, Mexico. Both small-sized raptors are considered in the same trophic guild, and as keystone species, interactions between them could be of interest for wildlife management in fragmented landscapes for agricultural activities. The diet analysis of these top predators could be a good monitor of prey abundance, some of which may be considered detrimental for agricultural activities. Prey frequencies in owl pellets were higher for arthropods (53.1%) and rodents (32.4%), but rodents were the most abundant item in terms of biomass (34.5%). Arthropods also showed higher relative frequencies in kestrel pellets (36.4%), but reptiles were the second in order of importance (28.4%) and contributed the most (64.7%) to the total prey biomass. The biomass contribution (66.1%) of vertebrates in the diet of the Burrowing Owl was lower than for the American Kestrel (95.6%) in a fragmented habitat area. Most prey types were present in the diet of both raptors, but we found significant differences in the biomass contributions of each category. Reptiles and rodents were the prey items that contributed most to the differences in the raptors’ diets, in terms of biomass (33% and 20.3%, respectively). Close nesting surroundings (0.5 km and 1 km radius) evidenced a higher proportion of natural vegetation in the kestrel’s (50.6% and 38.9%, respectively) than in owl’s potential home range (25.8% and 16.4%). The differences in diet suggest some degree of niche partitioning of these species likely due to the more flexible owl’s nesting habitat requirements and to the extended nocturnal activity of the species. These results reflect the capability of being complementary species, both exploiting some pests that are harmful for agriculture, and evidencing the importance of both species as regulators of agro-ecosystems in the peninsula of Baja California.
Full-text available
Transboundary frontiers often contain wildlife habitats that are fractured by geopolitical borders, which could have adverse effects on the wildlife that inhabit those areas. We examined the movement, breeding, roosting, and diet of 15 GPS-tagged barn owls (Tyto alba) along the Israeli-Jordanian and Israeli-Palestinian Authority borders. Our results showed that 80% of the barn owls hunted across the border. On average, the owls engaged in 5.4 hunting trips per night, 16% of which were cross-border excursions, and they crossed the borders as frequently as expected randomly, highlighting the importance of wildlife protection on all sides. Hunting movement, rather than cross-border activity, had an effect on the annual reproductive success and diet composition. Specifically, female owls that spent more time at the nest and engaged in longer distance hunting trips had higher reproductive success. Eighty percent of the females roosted outside of the nest box at a median distance of 908 m (range = 199–4112 m). Only 13.3% of the owls (2/15 owls) roosted across the border. These results increase our understanding of the movement, breeding, and roosting behavior of a non-migratory avian raptor. However, a serious lack of cooperation and communication between bordering countries hampers our ability to understand the full effects of differing environmental policies on a species that knows no borders.
Biological control of rodents in agricultural areas, increasing avian predator abundance by nest-box provisioning, has been proposed during the last decades in several regions around the world as an alternative to the widespread use of anticoagulant rodenticides (AR) to protect crops. However these experiences have often lacked a “Before-After-Control-Impact” (BACI) experimental design providing evidence strong enough about their efficacy reducing vole abundance or crop damages. Furthermore, little is known about possible indirect effects on non-target species. Here we test the effectiveness of providing nest-boxes to common kestrels (Falco tinnunculus) and barn owls (Tyto alba) in reducing the abundance of two different vole species using indirect index methods to estimate rodent abundance in late spring (May). We monitored the abundance of both vole species in a treatment (with nest boxes) and control (without nest boxes) areas during three years in alfalfa fields, using a BACI design. We studied also the efficiency of this control method in three different crops (alfalfa, cereal and fruit-trees) during the last year of the experiment. We compare the results obtained in this study area in NE Spain with those obtained simultaneously in other three experimental areas in NW Spain. Finally we explored potential negative effects of this method on the abundance of several passerine birds present in our research area. Our results showed a clear-cut reduction in the abundance of the two vole species, the common vole (Microtus arvalis) and the mediterranean-pine vole (Microtus duodecimcostatus) in the experimental area after applying the treatment (nest box installation) in alfalfa crops. We also found a significant decrease in the presence of both species of voles on fruit-tree plantations and alfalfa crops but we did not find significant effects in the cereal fields, where voles were in general very scarce. These results suggest higher efficacy of biological control in this study area than in some areas of NW Spain, what could be explained by several non-exclusive hypotheses presented in the discussion. For other species, we found a negative effect of nest-boxes on the abundance of a limited number of common bird species, but also positive effects over other species using poles with nest-boxes as perches. In conclusion, under the conditions of our study area, nest-box provisioning seemed to be an effective technique to reduce vole abundance, but also had some undesired side-effects on non-target species. However, these side-effects affected mainly to common bird species and can thus be considered less relevant than those caused by SGARs, often affecting to endangered rodent predators and that could thus have negative effects over any potential future natural control of vole abundance, potentially increasing future needs of SGAR use in a vicious circle.
Full-text available
Environmental hazards presented by pesticides and other toxicants are determined by exposure as well as toxicity. Predicting and estimating exposure is much more difficult than measuring toxicity since exposure depends on toxicant distribution, individual behaviour and population dynamics. We present a generalized compartment model for ecotoxicology and illustrate its use in relation to rodenticide hazard evaluation. The model highlights the need for particular sorts of field experiment. We present the results of a series of field experiments designed to assess potential effects of rodenticide use on some non-target animal populations.
Full-text available
To date, nest orientation and location in hole-nesting birds have been studied mainly in temperate regions and in diurnal cavity breeders. Here we studied the effect of exposure, orientation, and habitat on nest box occupation and breeding success of Barn Owls in a semi-arid environment. The occupation of nest boxes varied with exposure and orientation. A higher percentage of occupation and more Barn Owl nestlings per breeding attempt were found in nest boxes located in the shade than in the sun, and in those facing east/north rather than other directions. The temperature in the nest boxes varied, being lowest in those located in the shade and in those facing east. Nest boxes located in crop fields fledged more young per breeding attempt than those located in date plantations. We suggest that the higher nest box occupation and number of nestlings fledged was probably due to the lower temperatures in those boxes, an important factor in a hot/arid environment, although alternative explanations are also considered.
A guide to using S environments to perform statistical analyses providing both an introduction to the use of S and a course in modern statistical methods. The emphasis is on presenting practical problems and full analyses of real data sets.
Identifies and exemplifies the consequences of human activity on animal life: hunting, habitat destruction, exploitation of the seashore, afforestation (together with fruit plantations and horticulture), veterinary care, poisoning, garbage (reduction of sites, but greater quantity and quality), changes in agricultural practices, and interbreeding with domestic conspecifics. -P.J.Jarvis
In this paper, a theoretically consistent approach to including time costs in recreational demand models is developed. The demand model is conditional on the recreationist's labor market situation. For individuals at corner solutions in the labor market, utility maximization is subject to two constraints, leading to a demand function with travel costs and travel time as independent variables. With interior solutions in the labor market, time is valued at the wage rate and combined with travel costs to produce one 'full cost' variable. In an illustration, welfare measures based on the new model are estimated for a sample of sportfishermen. -Authors