Front Cover
 Center information
 Biological background
 Economic issues
 Rate equations
 Damages under time discounting
 Empirical estimation
 Expected damages based on the CLIMEX...
 Opportunity cost of quarantine...
 Sensitivity analysis and concl...
 Tables and figures

Group Title: Working paper - International Agricultural Trade and Policy Center. University of Florida ; WPTC 04-08
Title: Economic impacts of pink hibiscus mealybug in Florida and the United States
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00089793/00001
 Material Information
Title: Economic impacts of pink hibiscus mealybug in Florida and the United States
Series Title: Working paper - International Agricultural Trade and Policy Center. University of Florida ; WPTC 04-08
Physical Description: Book
Language: English
Creator: Ranjan, Ram
Publisher: International Agricultural Trade and Policy Center. University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2004
 Record Information
Bibliographic ID: UF00089793
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.


This item has the following downloads:

wp040008 ( PDF )

Table of Contents
    Front Cover
        Page i
    Center information
        Page ii
        Page 1
        Page 2
        Page 3
    Biological background
        Page 4
        Page 5
    Economic issues
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
        Page 11
    Rate equations
        Page 12
        Page 13
    Damages under time discounting
        Page 14
    Empirical estimation
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
    Expected damages based on the CLIMEX model predictions
        Page 21
    Opportunity cost of quarantines
        Page 22
    Sensitivity analysis and conclusion
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
    Tables and figures
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
Full Text

WPTC 04-08

i -ional Agricultural Trade and Policy Center

Ram Ranjan

WPTC 04-08 December 2004




Institute of Food and Agricultural Sciences




The International Agricultural Trade and Policy Center (IATPC) was established in 1990
in the Institute of Food and Agriculture Sciences (IFAS) at the University of Florida
(UF). The mission of the Center is to conduct a multi-disciplinary research, education and
outreach program with a major focus on issues that influence competitiveness of specialty
crop agriculture in support of consumers, industry, resource owners and policy makers.
The Center facilitates collaborative research, education and outreach programs across
colleges of the university, with other universities and with state, national and
international organizations. The Center's objectives are to:

* Serve as the University-wide focal point for research on international trade,
domestic and foreign legal and policy issues influencing specialty crop agriculture.
* Support initiatives that enable a better understanding of state, U.S. and international
policy issues impacting the competitiveness of specialty crops locally, nationally,
and internationally.
* Serve as a nation-wide resource for research on public policy issues concerning
specialty crops.
* Disseminate research results to, and interact with, policymakers; research, business,
industry, and resource groups; and state, federal, and international agencies to
facilitate the policy debate on specialty crop issues.

Economic Impacts of Pink Hibiscus Mealybug in Florida and the United

Ram Ranj an1
Postdoctoral Associate
International Agricultural Trade and Policy Center
Food and Resource Economics Department
University of Florida
Email: rranjan(aifas.ufl.edu
Ph: 352 392 1881-326
Fax 352 392 9898


This paper estimates the direct and indirect impacts of the Pink Hibiscus Mealybug
infestation on the economies of Florida and the rest of the United States. The approach
involves a Markov chain analysis wherein both short run and long run expected damages
from infestation are calculated. Use is made of the CLIMEX model that predicts the
potential pest-establishment regions in the US. While predictions based upon the
CLIMEX model extend the scope of damages beyond Florida, the damages are dependent
upon the rate of arrival and detection of species in those regions. Damages are
significantly higher when a longer time horizon is considered. When nursery owners
bear the full cost of quarantines in the form of loss of sales and treatment costs of
infected plants, the cost-effectiveness of quarantines as a regulatory tool is diminished.

Presented at the Annual International Agricultural Trade and Policy Center Conference,
December 7-8, 2004

1 Dale Meyerdirk, at the USDA-APHIS, graciously provided data related to the biological aspects of the
PHMB and information on the CLIMEX model predictions, besides making available several articles on
PHMB and commenting on this draft. Richard Clark with the Dept. of Agriculture and Consumer Services
(DOACS), Florida provide data related to the quarantines in various counties of Florida. Help is also
acknowledged from several others including Divina Amalin at APHIS, Florida, and Ed Bums with the
DOACS, Florida.


Invasive species management requires active participation of policy makers at various

levels. Monitoring at ports of entry for prevention, inspection and quarantining of

infested areas, biological, chemical and physical control are some of the several

management options available to the policy makers. However, the implementation of

such options is often done on an ad hoc basis and without considering the possibility of

their effectiveness in terms of costs, risk reduction or damage mitigation. One specific

example is the use of quarantines in order to prevent further spread of pests from an

already infested region. Quarantines are a useful means of preventing pest-spread, but

their effectiveness is limited by the modes of transport of the pest, number of ports of

entry for the pest and availability of alternative means to control the pest at lower costs.

For instance, certain pests can be kept under control through the use of biological agents

at a much lower cost than trying to prevent their spread through costly quarantines.

However, the application of quarantines is often guided by tangential objectives such as

stemming the decline in trade in an infested region caused by adverse reaction to pest


One such pest that underscores the above point is the Pink Hibiscus Mealybug

(PHM). PHM has already arrived in southern region of Florida (and some other

territories of the US), but has been kept under control due to an early and efficient use of

biological control agents. However, it has not been eliminated and will continue to be

considered a secondary pest under biological control, with new cases occurring every

now and then1. As a consequence, policymakers have to invest significant resources

towards minimizing their spread through monitoring and control. Private resource owners

too incur substantial costs due to imposition of quarantines and mandatory treatments of

infested plants. Considerable threat exists that the PHM will spread in to the rest of the

US, thus increasing overall costs significantly. The overall annual cost of control and

damages to the US economy from PHM have been estimated to be US $700 million, with

the global total being about $5 billion (Mofitt 1999, ARS 2003). One study puts the

economic costs of PHM invasion to US agriculture alone at $750 million/year when no

control measures are taken (Mofitt 1998). PHM infestation outside the US has caused

high agricultural losses. The agricultural losses to Grenada and Trinidad (in absence of

control measures) in the first year of introduction of PHM have been estimated to be US

$10 and $18 million respectively. Current economic losses exceed US $3.5 million per

year in Grenada and US $125 million per year in Trinidad and Tobago (USDA-APHIS

2003). Whereas, in Puerto Rico this species was detected early on and biological control

measures were employed, thus avoiding any agricultural losses (Michaud, 2002).

This paper estimates the current and potential costs of PHM infestation and spread

to the economies of Florida and United States. These estimates, however, are based

under the assumption that the regulator follows an 'optimum' policy of imposing

quarantines in detected regions and releases biological control agents at all PHM

infestation sites. A Markov chain framework is developed that incorporates the

uncertainties associated with the biological (such as arrival and spread of species) and

policy parameters (such as detection of infested regions) in order to calculate the

expected economic damages, both in the long and the short run. Use is made of

CLIMEX model predictions of the potential regions in the US favorable to this insect's

establishment. Finally, Numerical simulations are performed and key policy issues are

taken up in light of their findings.

This study would contribute to the literature on invasive species in several

regards. First of all, the case of PHM is unique as it has hosts spanning more than 250

species, a large number of which are agricultural commodities of significant economic

value. Findings from this study could be directly applicable to other invasive species

affecting similar hosts in future. Second, the PHM has been detected only in parts of

Florida and California, and is yet to spread into the rest of the United States. As a

consequence, significant effort is being dedicated towards containing further spread of

PHM through quarantine measures. By comparing the effectiveness of quarantine

measures on rates of spread of PHM to the costs of such measures, this study lays out

scenarios under which such policy measures could be justified. An indiscriminate policy

of quarantining every infestation may provide perverse incentives to affected businesses

and reduce its effectiveness by inducing under-reporting of infestations. Finally, this

study also points out the long run implications of pest infestations by considering all

possible scenarios of spatial infestation. Use is made of scientific predictions for

ascertaining scenarios.

Biological Background

The PHM (native of India), first reported in Egypt in 1920, was introduced to the island

of Grenada in the Caribbean in 1993. It has currently spread to 27 Caribbean islands. Its

primary host is the Hibiscus spp. on which it rapidly grows into colonies and is believed

to inject a plant-toxin causing severe distortions to the plant parts. Overall, it can affect

more than 250 species of plants which include coffee, guava, citrus, grape, peanuts, rose,

beans, coconuts, maize, sugar cane, soybean, cotton, etc. It is also found in regions of

Africa, Middle East, India, Pakistan, and South East Asia (USDA and APHIS 2003). In

the past it has led to a loss of up to 100% of agricultural output (grapes, jute, sorrel, etc.)

in India. It is also found in Hawaii, but its effect has been minimal there due to the

presence of its natural enemies.

Both sexes of the species are about 3 mm long. The average life cycle spans 45

days depending upon the temperature. A female can lay more than 500 eggs at one time.

Identification of the bug is not easy and can be positively done only by a taxonomist.

Modes of transport include crawler and egg sack dispersion through wind and by

movement attaching or sticking to animals or transported objects. Nursery plants and

trade of infested commodities also lead to its spread. Sometimes, ants that are attracted

to its honeydew may act as protectors and movers of PHM.

A number of biological control measures such as parasitoids have been employed

to control this invasive species with high success rates. Parasitoids grow inside the body

of PHM and eat it internally, eventually leading to its death. One particular parasite,

Anagyrus kamali, has been found to be very effective against the PHM. A generalist

predator, the red headed ladybird beetle, Cryptolaemus Montrouzieri has been shown to

be effective in controlling the PHM. A single ladybird beetle can kill about 3000-5000

Mealybugs in its lifetime. However, these may interfere with other biological methods

like Anagyrus kamali by sucking on the parasitized PHM. While ladybird is considered

a short-term solution to the PHM, parasitoids are the long-term solutions (USDA and

APHIS 2003).

The biological parameters of the PHM and A. Akamali are compared in Table 1.

Figure 1 shows the time paths of PHM and A. kamali. Notice that A. kamali takes over

the PHM population within a short span of 10 days even though its starting population is

one tenth of the PHM's starting population.


Though the biological control methods have been found to be very effective, they

will not lead to eradication of the PHM. As a consequence, biological methods may need

to be combined with other measures to ensure maximum safety. Most pesticides have

been found to be ineffective due to wax like secretion on the PHM's body, which cannot

be easily penetrated (USDA and APHIS 2003). However, Zettler et al. (2002) find that

post harvest treatment of PHM-affected crops with Methyl bromide leads to 100%

mortality of the PHM at all stages. Methyl bromide may adversely affect the quality of

the treated crop and as a result is used selectively on certain crops.

The PHM does not directly harm humans. The biological agents too have been

argued to be harmless. There has been no non-target impacts of the parasitoids used

against PHM to date.

Economic Issues

This paper models the economic impacts of PHM infestation by incorporating the

damages from pest infestation together with the costs of management options, such as

quarantines, into a stochastic framework that considers the risks of pest infestation and

spread. The paper, however, does not seek to optimize with respect to the costs and

benefits of PHM management. Instead, it takes the current management strategies as

given and considers the long term implications of such strategies on PHM spread and

subsequently on the economy. This approach is influenced by two main considerations.

First, the management options are currently limited to control measures due to the fact

that PHM has already arrived in Florida and some other parts of the US. Second,

biological measures of control are highly effective, but they cannot fully eliminate the

pest. As a consequence, quarantine measures are being combined with biological

measures to prevent its further spread. Current management strategy allows for limited

variability in the use of either biological methods or quarantines. Therefore, the key issue

is to consider the cost-effectiveness of such measures as the pest spreads. A stochastic

analysis of the pest spread and its damages (influenced by control measures) would help

guide PHM management in the long run.

The economic impacts of PHM can be classified into direct and indirect. Direct

impacts include the costs of prevention, control and monitoring besides the damages to

the host species. The indirect impacts include loss in businesses from quarantine, loss in

trade from supply disruptions and non-tariff barriers to prevent the arrival and spread of

the pest. Most studies on economic impact of invasive species fail to adequately

incorporate these indirect impacts, which could overwhelm the direct impacts. In this

paper the indirect impact from quarantines is considered explicitly as a part of the overall

damages from PHM.

Tables 2 through 6 below calculate the direct annual economic losses from PHM

infestation. The estimation procedure is based upon an earlier work by Moffitt (1999)

where the economic losses to key agricultural hosts of the PHM were calculated based

upon expert predictions of the damages to hosts in the event of no control being taken.

Using the same estimates of the proportional losses to hosts such as Avocadoes, Cotton,

Citrus, Soybean, vegetables, peanuts and Nurseries, economic losses are recalculated.

While these estimates give a rough account of potential damages caused by the PHM, a

much more detailed analysis is required to understand the threat from this pest both in

terms of its spread probabilities using scientific information and incorporating the

indirect economic losses.

INSERT Tables 2-6 HERE

In order to make more scientifically informed calculation of the potential damages

from the pest we make use of the CLIMEX model predictions of the degree of infestation

of PHM in the United Sates. The CLIMEX model was developed by the Commonwealth

Scientific and Industrial Research Organization and Cooperative Research Center for

Tropical Pest Management, Australia. This model uses PHM-infested regions in the

world that resemble the climates at various locations in North America to predict the

possible establishment of the PHM. Two predictions are available based upon 'match

levels' of 0.5 and 0.6. The match levels are based upon the climatic similarity of

locations under study in the CLIMEX model to the regions in North America (USDA-

APHIS 1998). A match level of 1 would imply that the climate of the target location

matches perfectly with the climate of the region where the infestation has taken place in

the past. Based on this ranking, a point 0.6 match level can be understood to have a

higher predictive capacity. At 0.6 match level, eleven sensitive States in the US were

identified as potential locations for PHM infestation. These are: Alabama, Arizona,

California, Florida, Georgia, Louisiana, Mississippi, New Mexico, North Carolina, South

Carolina, and Texas. At 0.5 match level potential States are: Alabama, Arizona,

California, Florida, Georgia, Arkansas, Louisiana, Maryland, Mississippi, Nevada, New

Mexico, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, and Virginia.

Using a 0.6 match level, economic losses are estimated and presented in Tables 2

through 6 above. Notice that there is a significant reduction in predicted damages after

using CLIMEX model forecasts.

Next we model the indirect impact of PHM infestation such as loss in business

from quarantines, treatment costs etc. We model the total economic impact of PHM

through a Markov chain analysis. Since the pest is under control in the Florida region,

the direct economic damages are minimal. However, there has been a continuous arrival

of new pests in the various counties over the last three years. Figure 2 shows the arrival

sequence into various counties of Florida.


Due to these constant arrivals, regulatory agencies such as the USDA enforce quarantines

on the infected regions. These quarantines are mostly imposed upon nurseries, as

Hibiscus (a nursery plant) being the primary host of this insect is the first one to be

infected. There is a significant cost to the nurseries from loss of revenues during the

quarantined period besides the costs of treating infested areas. Several nurseries have

gone out of business due to such quarantines in the past years2. Figure 3 shows the

composition of Nurseries in terms of their annual revenues. Yet, the significance of such

quarantines cannot be overemphasized in terms of reduction in risk of spread outside

Florida into the rest of the US.


The approach adopted in this paper is to model the processes in the PHM

infestation (such as arrival, spread, re-infestation, etc.) and regulatory reactions (such as

quarantines) as a continuous time Markov process. A continuous time Markov process

assumes that the rates (of arrival, spread, detection, etc. of pests) follow an exponential

distribution. That is, a process shifts from one state of the system into another after an

exponential amount of time. Such processes have been commonly used to describe

biological phenomenon such as the birth and death rates of species. Parameters related to

pest infestation have been modeled as emanating from a Markov process in the past

(Zimmerman 2002). Markov chains have also been highly successful in mimicking

various societal phenomenon such as labor migration, population distribution, traffic

movements etc. One major advantage of such an approach is that it offers convenience of

empirical estimation and transparency of analysis.


The model below delineates the US region into two parts, FL named as region A and rest

of US, named as region B. There are three main 'states' possible for these regions,

namely; un-infested (u), infested (i) and under quarantine (q). Given these three main

'states', the possible state space is a nine by one matrix as shown below:

{A, B, A, B, A, B, A B, A, B,A, B, AqB, AqB, AqBq

These states capture the various possible combinations that are possible between the two

regions3. For instance, A,,B, refers to the state when both the regions are free of any

infestation and AqB refers to the state when Florida is in the state of quarantine and rest

of the world is un-infested4

Key Parameters:

The important parameters of concern are the arrival rates of PHM from an outside region

into Florida and the rest of US (ae, be), the rates of infestations from one region into

another (a,,b,), rates of detection of an infestation (da,db), rates of disinfestations of

infested regions due to control measures(Sa, b), rates of disinfestations of quarantined

regions (3aq, 8bq), and the rates of re-infestation of the quarantined regions (r, rb). All

rates are defined in terms of units per year and are detailed below:

Arrival rate into regions B and A (be ae) : The arrival rates are defined as the number

of arrivals of the pest per unit of time. In certain cases it may not be possible to measure

the exact stock of pests that arrive at the point of interest or relate that stock meaningfully

to economic damages. An alternative measure in such cases would be to relate the arrival

rate to certain observable parameters such as number of observations of pest infestation

over a certain period.

Infestation rate from A to B and from B to A (a,, b,): Infestation rates between two regions

are defined as the number of transmissions of pest from one infected region to another

within a certain time period. For practical purposes, these transmissions could be

measured by the number of detections of infested shipments from one region to another.

Rate of disinfestations from a region due to bio-control (8', 8b): Rate of dis-infestation

is defined as the time it takes for pest to be eradicated from a certain region.

Alternatively, it could also be measured as the number of disinfestations per unit of time.

Rate of dis-infestation after quarantine (3, 6,qb): It is possible for the quarantined

regions to be dis-infested at a different rate as compared to infested regions that are not

yet quarantined.

Rate of re-infestation of a quarantined region (ra, rb): This parameter incorporates the

possibility that quarantined regions may fall back into a state of infestation instead of

getting dis-infested after the quarantine is removed.

Rate of detection of infested regions and fall into quarantined states (da, db): This

measures the rate at which infested regions are detected and placed under quarantine.

These rates define the transition process from one state of the system into another.

For instance, when the arrival rate of species into Florida is higher than that into the rest

of the US, the likelihood of finding states when Florida is infested as compared to those

when the rest of the world is infested would be higher over a given time horizon. Given

such rates, it is also possible to find the long term behavior of the system, which is of

special interest to us as it would throw light on the economic aspects of pest infestation in

the long run. The transition diagram is shown in the figure below.


Rate Equations

In order to solve for the long-term behavior of the system, one needs to look into the

steady state behavior of the system. The steady state is derived from the fact that in the

long term, the net arrival out of any given state must equal the net entry into it. Using

this, we derive the first nine of the below equations. In these equations, P (with

subscripts) represents the long-term probability of finding the system in that state. This

term can also be interpreted as the fraction of time spent in that state in the long run. The

last equation (10) is derived from the fact that the sum of the fractions of time spent in all

possible states must equal one.

(1) Pu (b, +a,) =P,,b +Po +P q aq +P qbq

(2) P, (db +b, +a, +Sb P ,r P,,be + P, f + Pq,aq

(3) Pq (rb +a, +b, + bq) P, db + Pl o +P qaq

(4) P.(a, + b + d + ()= P, b +Pqra + P.a, +Pq fbq

(5) P, (db+ + b, + () =P. (a, +b,)+P,,(b, +a,)+Pqrb + P, r

(6) Pq (rb< d, + + bq= Pdb +Pq (a, + b,) +Pqr

(7) Pq (r + a, + be + ') = P,,u + Pq, b + Pqqbq

(8) Pq, (b + r +db + ) aq= Pu (a, + be) + P,d + Pqr,

(9) Pq (ra +rb +6bq+ +aq) Pq,db + P,qda

(10) P + P, + P, + + P, + P + P,, + P +P = 1

Solution of these rate equations would yield the steady state probabilities P. Once the

fraction of time spent in each state is derived, the economic analysis is fairly

straightforward. For instance, if one is interested in solving for the expected damages in

the long run, given the above characteristics of the system, the analysis would involve

multiplying the economic damages in each of the states by the fraction of time spent in

each state as:

(11) ZD(P )*P

Damages under Time Discounting

The above presented a way to calculate the expected sum of damages from various

possible states of PHM infestation over a year. However, one key question of concern

may also be the expected sum of damages over a longer period of time when the planner

may have time preferences. It is pertinent to note that when the expected damages are

taken over a longer time horizon, the current state of infestation may have an influence

over the total sum. That is, the sum of expected damages would vary depending upon

whether one started in AB, or AqBq. This is because, each state of the system has a

unique steady state rate of departure and entry that may be different from the others. In

order to calculate an infinite horizon sum of damages, we define g(xy) as the sum of

damages if one started in state x for Florida and state y for the rest of the US. Following

the derivation of average expected discounted costs in Kulkami (1995), the relation

between the generator matrix (Q), per period payoffs in each state and the long run

expected profits from starting in each state can be derived as follows 7:

gu~ D(P,)

guq D(Pq)
g,, D(P, )

gqu D(Pq )
gq, D(Pq,)
_gqq D(Pq)_

The generator matrix Q, which is basically Figure 4 in the matrix form is derived below.

The right hand side denotes the per-period damages in each state.


Note that the diagonal elements, which are marked with stars, represent the negative sum

of all rates in that row. For instance, the first row represents the departure rate out of the

state AB, into all other states. The elements of column one and row one seek to balance

the departure rate out of this state.

Empirical Estimation

Estimation of the key parameters of the model (such as the arrival and spread rates of the

PHM) is no mean task even for a simple model like this. There has been little scientific

work done to get estimates of the arrival and infestation rates of this species. Most of the

work is currently focused upon surveying the impact of biological control agents on the

PHM. There have been some observations of PHM behavior under simulated conditions

in the laboratory that have yielded the growth and survival rate for PHM and its

biological control agents such as A. Kamali. However, at this stage there is limited

information available with respect to the specific interaction between the PHM and its

innumerous hosts. For a more detailed modeling approach, one would require

information such as the density of PHM species on each host plant and the variance of

this density in presence of multiple hosts. As a consequence, our estimates of the various

rates are based upon some simplifying assumptions. The arrival rates, rates of detection

and quarantine, re-infestation and disinfestations rates are all calculated from data

available on quarantine of nurseries in the ten counties of Florida so far. Since the

infestation has not spread beyond Florida, hypothetical estimates of the same rates are

proposed. Below is presented a brief account of derivation of these rates.

The arrival rate into Florida is based upon the assumption that the number of

detections in various counties of Florida was each independent arrival from an outside

region. Further assuming that the arrival rate was a Poisson process, the average arrival

rate into Florida (ae ) was estimated to be 12.33 per year. Since there have been no

known infestations into regions outside Florida (except California and Hawaii), we

assume a very low arrival rate from outside into the rest of the US (be=.001). Ideally,

infestation should be defined in terms of the pest reaching some critical observable

threshold. However, practically, infestations are recognized only after detection on some

private property or in nurseries. Consequently, infestation and detection rates are treated

as same in this paper.

Assuming that Florida is one of the first States to be infested, any further

infestations into the rest of the US can be deduced from number of infestations outside

the Florida region. Between 2002 and 2004, there has been only one detected case of a

shipment of infected nursery plants outside Florida. Given one such case of arrival

outside, one can assume the rate of infestation from Florida (region A) to the rest of the

US (B) to be 1/3. However, since there are forty nine such States, the arrival rate into

the rest of the US must be further divided by forty nine8 (a, 1 ). Due to no cases

from the rest of US into FL yet, we assume the rate of infestation from B to A as very

low (b, =.001).

An important point to note here is that the PHM is under control in Florida due to

the effectiveness of bio-control agents such as A. Kamali and others. However, there are

two important clauses to this; first it takes roughly one year for a new infestation to be

brought under 90-95 percent control (Amalin et al. 2003) and second, following the first,

it is not possible to eliminate the bug. From the first fact we can deduce the rate of dis-

infestation of the infected region to be 1( =1,,=b 1). The second fact emphasizes that

even dis-infested regions can fall back into a state of infestation.

A distinction needs to be made between disinfestations from states that are

quarantined and from states that are infected. While most of the hosts of the bug are

crops of significant agricultural value, the major host is the hibiscus plant, which is

grown in nurseries. It is significant to note that all detections so far have been made in

the nurseries, following which they were placed under quarantines. Quarantines, whereas

they reduce the chances of further spread, they also impose significant economic

hardships on the nurseries' revenues in terms of forgone sales, costs of treatment of

infected plants and even closure of businesses. In Florida, there were 575 nursery-days

of quarantines on 15 nurseries, whereas in 2003, there were 1008 days on 22 nurseries

combined9. This gives the average time spent by a nursery in quarantine as 0.12 years

per year. The rate of departure out of quarantine into dis-infestation is then given by the

reciprocal of the average time spent in the state of quarantine. From this we

derive: =8.67 b =8.67.

It is also possible that there is an instantaneous re-infestation of quarantined

regions after the quarantine is lifted. However, the data revealed a time lag before re-

infestation of the previously quarantined regions. Consequently, we assign low

possibilities to such events as: r=.001, r,=.001. Finally, we assume that all infestations

into nurseries are detected at the same rate as their arrival, giving us the average rate of

fall into quarantined states as: d, =12.33, db (1/3*49).

Note that the above estimation of parameters is based upon observations at a

disaggregated level of nurseries. It is possible that the rates of arrival, quarantine and

infestations outside Florida may differ when the problem is considered at a much

aggregate level of two regions. For instance, the rate of infestations outside of Florida

may be expected to be higher when the entire State is infested as compared to the case

when only a few counties are infested. Keeping such limitations in mind, we may

consider the above estimation to be the base case scenario. Next, we derive the steady

state fraction of time spent in each of the nine system-states as given below by the


PuP,,Pq .307 .0016 1.3X10 6
(13) PPPq =.283 .0015 1.2X10-6
P, p ,q .403 .0023 1.8X106

It is evident from above that the chances of infestation into the rest of the US are fairly

insignificant in the base case scenario. This is affected by our assumption of low

infestations out of Florida and from outside regions into the rest of the US. Also, the

system spends most time in the states when Florida is either un-infested, infested or

quarantined. These assumptions will have an impact on total expected damages

accordingly. Next, using the figures in Tables 2 though 6, we calculate the damages

from these various states, defined as the Dmatrix (in million US $)10

D =0
D,, =1,418.140
D,, =1,418.140
Dq = 4,374.894* 2
D, =162.856
D, =162.856+1,418.140
Dq =162.856 + 4,374.894*2
Dq =1,006.649*2
Dq, =1,006.649* 2 +1,418.140
Dq =4,374.894* 2 + 1,006.649*2

Note that while solving for the damages in the quarantine stages we multiply the loss to

businesses from quarantines by a factor of two in order to incorporate some of the

treatment costs. A brief telephonic survey revealed that nursery owners spent almost as

much as their monthly revenues over the treatment costs. Societal treatment costs, such

as release of parasitoids are much higher, however such costs are assumed to be

adequately covered in this doubling of the quarantined costs.

The expected sum of damages to the entire US region in one year is, simply, the

sum of the product of elements in the Pmatrix with the corresponding elements in the

Dmatrix and equals US $871.787 million. Note that this figure is significantly lower than

the average annual damages of US $1,581 million as calculated earlier (as shown in

Table 6). This is due to the fact that the Markov model assigns lower steady state risks to

the rest of the US being in either the infested or quarantines states. Whereas, the earlier

estimate does not incorporate the risk-aspect of the PHM problem. In order to calculate

the expected discounted sum of damages over an infinite time horizon, we solve the

gmatrix for various states. The matrix of g's is derived for ten, five and one percent

interest rates respectively as (million US$):

gn 8,637 1,7350 8,7100
g, 9,918 1,8700 8,8500
guq 9,634 1,8360 8,8100
8,707 1,7420 8,7170
< 9,988 1,8770 8,8560
9,704 1,8430 8,8170
8,770 1,7490 8,7230
gqu 10,050 1,8830 8,8620
gq' 9,766 1,8490 8,8240
g qq (10 percent) percentn) percentn)

First thing to note here is that the long run damages are considerably higher than the

annual estimates Further note that the damages increase as the discount rate is lowered

from 10 percent to 1 percent. Also note that for a given discount rate, the highest

damages are felt when the system starts with quarantine in Florida and infestation in the

rest of the US. The least damages occur when the current state is of un-infestation in

both the regions, which is obvious. The states of quarantine cause high amounts of

damages, a result of incorporating the indirect economic impacts of the pest. Important

insights emerge from presenting the discounted sum of damages based upon the starting

state of the system. For instance, consider the results for the ten percent discount rate.

When the starting state is that of un-infestation in Florida, damages are higher with the

rest of the US being in the state of infestation as compared to it being in the state of

quarantine (compare g,, &gu). Whereas, when the starting state is that of un-

infestation in the rest of the US, damages are higher when Florida is in the state of

quarantine as compared to it being in the sate of infestation (compare g,, & g,, ). This

anomaly, is primarily due the long run propensity of the system to spend a high fraction

of time in the state where Florida is quarantined and the rest of the US is un-infested.

This has important policy implications as it warns against complacency. The fact that the

system is currently free from infestation is no indicator of the extent of damages in future.

It is possible that certain states may take a speedier transition to the most damaging states

as compared to others. The long run spatial distribution of pests is an important piece of

information to strive for, and management decisions based solely upon current state of

the system could be misleading. Therefore, besides understanding the magnitude of

resources at risk, it is also important to relate them to the long run risks in the chain of


Expected Damages based on the CLIMEX model Predictions

Using the 0.6 level predictions for potential establishment regions in the US we derive

the damage matrix as (million US$):

D, = 0,
D, = 675.047
Dq = 3883,
Di =162.856
Damage Matrix = Dii = 837.903
Di 4,045.8
Du = 2,013.3
Dq = 2,688.34
Dqq = 5,896.29

Taking the sum of product of the elements in damage matrix with the probability matrix

as above we get the expected sum of damages per period in the steady state as US

$867.580 million. Note that these damages are almost equal to the ones estimated above.

This is primarily due to our assumption of the system spending very little time in states

when the rest of the US is either infested or quarantined. As a consequence, the damages

captured here are still those in the Florida region. Finally, the total discounted value of

expected damages over an infinite time horizon for a 10 percent discount rate horizon is

derived as (million US$):

gu 8,599
gu' 9,209
guq 9,041
g,, 8,669
g =, = 9,278
gq 9,111
gqu 8,731
gq 9,340
g [99,173

Opportunity Cost of Quarantines

As is evident from the steady state matrix of transition probabilities derived above, the

system spends most of the time in the state when Florida is quarantined. One crucial

issue is whether the costs of quarantine are worth their utility. We do not really know

what kind of infestation rate we would get into the rest of the US if the quarantines were

not imposed into Florida nurseries. Assume that the current rate of infestation from

Florida into the rest of the world a,=1/(3*49) is a result of the stringent quarantine efforts.

Also assume that in the absence of quarantines the rate of arrival into Florida will equal

the rate of departure out of Florida and into the rest of the US (a,=12.33). In such a case

the annual impact to the overall economy of the US when no quarantines are imposed,

can be derived by taking a product of the revised damage matrix with its long run steady

state probabilities. Note that the revised damage matrix would have zero damages in the

states of quarantines for either of the regions. Following the above approach, the

expected annual damages are derived to be US $205.697 million. The impact on the US

economy in the presence of quarantines is the base case scenario derived above as US

$867.580 million. Therefore, taking the difference between the two we find that the

opportunity cost of quarantines is actually a positive number equal to US $661.883

million. This extra cost of quarantines can only be justified if either the damages are

expected to be much higher than assumed above or if the risks of spread are greater.

However, the actual cost of quarantines may itself be lower if businesses do not suffer

complete loss of sales during the quarantine period as assumed here, or if the treatment

costs which are included as a part of quarantines are much lower. In the above

simulations it was assumed that the treatment costs of infected plants in the nurseries

were equal to the loss of sales, thus doubling the quarantine costs. In the following

section we relax some of the above assumptions to gain further insights.

Sensitivity Analysis and Conclusion

Using the above CLIMEX predictions as the base case we perform some simulations to

study the impact of variations in our assumptions over the key parameters. The results of

these simulations are presented in Table 7 below.


Case 1 looks at a case when the rest of the US, as predicted by the CLIMEX model, have

the same escape rate as the arrival rate into Florida. Further, the rate of detection in the

rest of the US remains as before. That is, a, = 12.33, db = 1/(3 49). Damages increase

significantly after this manipulation. However, when the rate of detection is increased to

a high level, equal to the rate of arrival, (case 4), the damages are almost three times

higher as compared to the base case. Notice that quarantines have a large impact on the

damages and therefore must be justified in terms of their impact on future risk reduction.

With increasing susceptibility of the geographical region, either due to trade or

exogenous reasons, the arrival and spread rate of species may not show any linear

relationship to quarantines beyond a certain threshold. That is, beyond a certain point the

effectiveness of quarantines fall whereas their costs may rise. Therefore, it is significant

to know the relation between the impact of quarantines on future risk of pest spread and

consequential damages in order to justify their costs. Case 2 leads to lower damages than

the base case as the control measures are twice as effective leading to higher rates of

disinfestations. Damages are twice as high as the base case when the arrival rate into the

rest of the US from outside is increased significantly. Finally, when the costs of

quarantine are reduced to half, by restricting to loss in revenues only, the expected

damages are reduced to half their value from the base case. However, this number is still

twice as much as the case when quarantines are not imposed at all, as shown above in the

calculation of the opportunity cost of quarantines.

At this stage the paucity of data does not allow us to put our bets on any of these

numbers derived above, but the simulation analyses do help throw light on the merits of

regulatory policies such as quarantines. It is evident that that there is a limit to which

such measures can be effective. Beyond a certain point when the arrival rate of species

increases due to exogenous reasons, or when the costs of preventing arrivals increase, it

would be wise to take recourse to alternative ways of pest management such as direct

control. The main findings of the paper are not the high economic damages from PHM

infestation, but the fact that high damages could itself be partly caused by the 'optimal'

management procedures such as quarantining every case of detection, unless care is taken

to consider the cost-effectiveness of such policies. It is also important to allocate policy

measures based upon the long run impacts rather than a short-term horizon, as the

damages from the pest are dependent upon the spatial distribution of pest in the long run.


1 In 1999, it was found in the imperial county of California. In 2002, the PHM was

located in Broward and Miami-Dade counties of Florida. By the end of 2004, more than

10 counties in Florida were reported to have PHM infestation.

2 While regulators make an effort to restrict the impact of quarantines to the sale of the

infested plant, the actual impact depends upon the severity of infestation and the number

of host plants infested. Communication with the affected nurseries has revealed that this

impact could range from partial to entire loss of revenues during the period of

quarantines. In this study, it is assumed that quarantines lead to a total loss of revenues.

3 The 'states' of the system should not be confused with the fifty 'States' in US.

4 'Florida being in a state of quarantine' is a figure of speech. It is possible that multiple

states such as quarantine and infestation exist in the same region, and is a function of the

level of dis-aggregation assumed within a region. For instance, if quarantines are placed

solely on nurseries (which is the case now) it is possible to classify the states as has been

done in the paper. When quarantines are placed also on the agricultural sector, the state

space would have to be enlarged and states redefined.

5The estimation of the key parameters was based upon past data on quarantines on

infested nurseries in Florida. This data was provided by the Florida department of

Agriculture and Consumer Services, and is available on request.

6There are two more arrows connecting states AqBto A ,Band A,,B to AB, that have been

omitted in the above figure to maintain presentability.

7See Kulkami (1995) pgs. 306-11 for more details

8At this stage there is no scientific information over the exact probability of arrival into

each of the individual 50 States. The division by 49 is done under the belief that the 49

States face a uniform chance of receiving the infested material. Further, no distinction is

made between arrival rate into Florida from outside and arrival from within due to

infested shipments.

9These exclude quarantines imposed due to re-infestations.

10Note that we need to subtract the values of Fl from the US to arrive at a rest of the US



1. Amalin, D. M. K. A. Bloem, D. Meyerdirk, and R. Nguyen. "Biological Control

of Pink Hibiscus Mealybug in South Florida: A One Year Assessment", USDA-

APHIS, Manuscript (2003).

2. ARS "On the Lookout for Scaly Invaders" (2003): URL


3. Clark, R. A., DOACS, IFAS, University of Florida List of Nurseries

Quarantined under the Pink Hibiscus Control Program Since 2002, Personal

Communication (2004).

4. Kulkarni, V.G. Modeling and Analysis of Stochastic Systems, Chapman and Hall

Publications, UK (1995).

5. Meyerdirk, D. E. and L. W. De Chi. "Models for Minimizing Risks of Dangerous

Pests: The Pink Hibiscus Mealybug and Papaya Mealybug" Proceedings of the

Caribbean Food Crops Society, Grenada. 39 (2003): 47-55.

6. Michaud, J.P. "Three Targets of Classical Biological Control in the Caribbean:

Success, Contribution and Failure", (2002), URL:


7. Moffitt, M. J. "Economic Risk to United States Agriculture of Pink Hibiscus

Mealybug Invasion", A Report to the APHIS, USDA under Cooperative

Agreement No. 98-8000-0104-CA at the University of Massachusetts, Amherst


8. Persad, A. and A. Khan "Comparisons of Life Table Parameter for M. Hirsutus,

A. Kamali, C. Montrouzieri and S. Coccivora", BioControl 47 (2002):137-149.

9. National Agricultural Statistics Service. Various Tables (2004):


10. Sagarra, L. A., and D. D. Peterkin. Invasion of the Caribbean by the Hibiscus

Mealybug, Maconellicoccus hirsutus Green (Homoptera: Pseudococcidae):

Phytoprotection. 80, (1999) : 03-113.

11. USDA-APHIS. "M. Hirsutus (Green): Simulation of Potential Geographical

Distribution Using CLIMEX Simulation Model", Internal Document, (1998).

12. USDA-APHIS (2003): (http://www.aphis.usda.gov/oa/pubs/PHMpaler.pdfT)

13. Zettler, J.L, P.A. Follett, R.F. Gill. "Susceptibility of Maconellicoccus Hirsutus

(Homoptera: Pseudococcidae) to Methyl Bromide" Journal of Economic

Entomology: vol. 95, No. 6, (2002): 1169-1173.

14. Zimmerman, K. M., J. A. Lockwood, A. V. Latchininsky "A Spatial Markovian

Model of Rangeland Grasshopper Population Dynamics: Do Long -Term Benfeits

Justify Suppression of Infestations?" Environmental Entomology, Vol. 33, No. 2,

(2002) :257-266.

Table 1: Biological Parameters for PHM and A. Kamali

Source: Persad and Khan (2002). Intrinsic rate of growth and finite
related as A = e'"' The doubling time of the species is defined as T =

rate of increase are
Ln(2) / rm

Table 2: Annual Average Value of Crops that are Hosts to the PHM ( in 2003 US$

Vegetables Avocado Citrus Cotton Peanuts Soybean Nursery
Florida 1,075,513 14,505 1,379,173 26,567 48,267 1,650 1,006,648.647
US 8,801,959 378,540 2,258,104 3,696,162 747,668 14,236,502 5,381,542.322
CLIMEX- 6,233,834 363,655.7 878,931.3 2,896,326 634,557.8 71,2246 194,197.853
*These are the eleven States predicted by the CLIMEX model, minus Florida
Source: National Agricultural Statistics Service

Table 3: Expected Fraction of Damages from PHM Infestation
Vegetables Avocado Citrus Cotton Peanuts Soybean Nursery
Florida .04 .3 .04 .01 .2 .2 .05
US .04 .3 .04 .01 .2 .2 .05
Source: Moffitt (1999)

Biological Parameters PHM A. Kamali

Intrinsic rate of growth .0801 .3301


Doubling Time (T) 8.63 2.09

Finite Rate of Increase 1.0834 1.39


A .

Table 4: Expected Average Damage in Dollar Amounts (in 2003 US$ 1000)
Vegetables Avocado Citrus Cotton Peanuts Soybean Nursery
Florida 43,021 4,351 55,167 266 9,653 330 50,332
US 352,078 113,562 90,324 36,962 149,534 2,847,300 269,077
CLIMEX 249,353 109,097 35,157 28,963 12,6912 28,490 97,075

Table 5: Expected Annual Damages from PHM by Moffitt ( in 2003 US$ 1000)
Vegetables Avocado Citrus Cotton Peanuts Soybean Nursery
Florida 40,587 3,400 66,958 240 58,537
US 214,095 72,937 104,176 43,025 247,383
*Not Considered.

Table 6: Total Annual Average Damages to Crops( in 2003 million US$)

Mottiff This Study
Florida 169,722 162.856
US 681,616 1,580.997
11-States, excluding Florida 675.047

Table 7: Sensitivity Analysis (in 2003 million US$)

Base Case 1: Case 2: Case 3: Case 4: Case 5:
Case a = 12.33, 3 = 2 b = 2 b =12.33 a =12.33 Cost of
db = 1/(3 49) d = 12.33 quarantines=1/2
of the base case

867.580 1,458.73 846.101 1,490.25 2,325.2 458.937

Figure 1: Time Paths of PHM and A. Kamali





10 C A.Kamali

No. of days

2 4 6 8 10

Note: Figure 1 compares the growth in stock over time for PHM and its predator A. Kamali. Note that the
population ofA.Kamali overcomes the PHM population injust ten days even ;ih. ,'ili its ',l,, r i stock is
much lower.

Figure 2: Sequnce of Infestation and Quarantine in Florida

10 10

0 10 20 30 40 50 60 70
Sequence of PHM detection in
various Counties of Florida
Source: Clark (2004)
Note: Figure 2 shows the pattern of infestation in various counties ofFlorida. The X-axis denotes the
order of detection over time and the Y-axis denotes the particular county in which PHMwas detected. For
instance, the first dot (numbered 1) represents the first case of detection in Broward county. The counties
are numbered as: 1-Broward, 2-Dade, 3-St. Lucie, 4-Brevard, 5-Palm Beach, 6-Pinellas, 7-Collier, 8-
Desoto, 9-Lee, 10-Hillsborough


Annual Value of Nurseries in Florida

7 21 48 88 109 145 233 709
Number of Operators

Source: Drawn from National Agricultural Statistics Service Figures
Note: Figure 3 matches the number of nursery operators in Florida with their annual value of sales in

Figure 4: Rate Diagram 6

-- 41A,, dA[,, B
A a




a, +b,




rb I
a qq4~


Note: In this figure the arrows demonstrate the linkages between two states of the system. There are nine
possible states of the system. However, it is not possible to move directly from any one state onto all other
states. The parameters reflect the rate at which this change in states is made possible. For instance, ra is
the rate at which the state of the system could change from A B. to AB, That is, region A turns into an
infested state at the rate ra per unit time, whereas there is no change in the state of region B during that
period. Alternatively, it can be stated that the system spends, on average, 1/ra amount of time in
A B, before moving to state AB

A -


a+ b ,

a, + be

a +b



> AqBi


Figure 5: Generator Matrix

A, B A, B A q AiB 4 AiB


AqB Aq B AqB
qu q2 q

be 0
* db

0 b, + a

b 0
0 0 *

oa 0 (b


0 0 (5a
S(aq 0 0
0 (aq 0
o 0,, o
0 0 S,,

0 (ae +b,)
+be 0
* db

0 0
ra 0
0 r

0 0 da
* be +a 0

Note: The generator matrix denotes the rate at which transition takes place between states. For instance,
the element (b, ) under the row AB, and the column AB, represents the rate at which rest of the US
gets infested by PHM arriving from regions outside the US. The elements marked star in any row are the
negative sum of departure rates out side the state represented by that row.





I q
A qBi
q q

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs