The Regional Economics Applications Laboatoy (REAL) of the Univesity of Illinois focuses on the development and use of analytical models fo uban and egional economic development. The pupose of the Discussion Papes is to ciculate intemediate and final esults of this eseach among eades within and outside REAL. The opinions and conclusions expessed in the papes ae those of the authos and do not necessaily epesent official positions of the Univesity of Illinois. All equests and comments should be diected to Geoffey J. D. Hewings, Diecto, Regional Economics Applications Laboatoy, 607 South Matthews, Ubana, IL, 61801-3671, phone (217) 333-4740, FAX (217) 244-9339. Web page: www.uiuc.edu/unit/eal DISCOSEM: AN INTEGRATED EPIDEMIOLOGICAL-ECONOMIC ANALYSIS OF FOOT AND MOUTH DISEASE IN THE SOUTHERN CONE by Kal M. Rich REAL 04-T-15 Novembe, 2004
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 2 DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 1 Kal M. Rich Depatment of Agicultual and Consume Economics, Univesity of Illinois at Ubana-Champaign, 427 Mumfod Hall, 1301 W. Gegoy Dive, Ubana, IL 61801, email: kalich@uiuc.edu. Intoduction Animal disease outbeaks pesent significant costs to affected counties, especially when the livestock secto is lage and substantially integated into intenational expot makets. Fo example, the ecent discovey of Bovine spongifom encephalopathy (BSE, o Mad Cow Disease) in cattle in the United States esulted in the immediate closue of almost 90 pecent of the U.S. expot maket fo beef. While the loss of access to expot makets may be bief in duation, animal diseases can also imply consideable expenditues in disease contol effots, indemnity payments fo destoyed animals, lost poduction, and losses in elated industies, including touism as in the case of foot and mouth disease (FMD) in Geat Bitain in 2001. Despite the economic impotance of animal disease outbeaks, thee has been elatively little wok to combine ealistic epidemiological models with sophisticated economic analysis. Because animal diseases (and poduction cycles) have paticula evolutions though time and space, the analysis should ideally be both spatial and dynamic. The impotance of the spatial component is often einfoced both by movements of animals and disease spead vectos acoss space. Meanwhile, time plays an impotant pat in animal disease contol analysis because of the dynamic and even stochastic natue of disease outbeaks and because of the ole of investment in livestock economics. In light of these issues, this pape develops an integated epidemiological- 1 This pape is based on peliminay esults fom the autho s doctoal dissetation, Spatial Models of Animal Disease Contol in South Ameica: The Case of Foot-and-Mouth Disease. The autho would like to thank Jose Bevejillo, Nick Bozovic, David Bullock, Peggy Caswell, Phil Gacia, Rodney Howe, Gay Mille, Cal Nelson, Lauian Unneveh, and Alex Winte-Nelson fo useful comments and suggestions thoughout the couse of the eseach. This eseach was funded in pat by the USDA Coopeative State Reseach Education and Extension Sevice (CSREES), Awad No. 2003-35400-12903.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 3 economic model of animal disease contol (DISease COntol Spatial Epidemiological-economic Model, DISCOSEM) tha in contast to the existing liteatue, is both dynamic and spatial. The model is intended to suppot analysis of altenative FMD contol policies in the Southen Cone. 2 The spatial component of the epidemiological model epesents an impovement ove past models of FMD (Beentsen, et al., 1992; Gane and Lack, 1995; Ekboi, 1999; Duand and Mahul, 2000), with the added advantage of being linked to a spatial economic model that can detemine the egional effects fom diffeent mitigation stategies. Moeove, the dynamic natue of the economic component is significantly diffeent fom past economic models of FMD contol and epesents a methodological impovement. Pevious patial-equilibium models have eithe been static (one-peiod) o shot-un analyses that measued the shot-un shocks to supply, poductivity, and expots (Schoenbaum and Disney, 2003; Mangen, et al., 2004). Othes have attempted to incopoate dynamics but have simply discounted the impact of one peiod ove time (Beentsen, et al., 1992). Since an FMD outbeak will engende changes in beeding decisions and input allocation in futue poduction, which in tun will have welfae impacts in the economy ove time, a dynamic appoach is suggested to captue these effects, as well as the dynamics of disease itself. Twelve diffeent contol stategies ae consideed based on the simulation esults of the epidemiological component applied to an initial outbeak in Paaguay. The model esults ae combined with exogenous costs such as vaccination and eadication costs, veteinay sevices, and othe govenment expenditues to detemine the total benefits and costs of an outbeak ove a five-yea peiod unde altenative mitigation stategies. Peliminay esults fom the model indicate that stamping out stategies have geate long-tem benefits than policies that utilize vaccination, given that stamping out policies engende shote dislocations on intenational makets than vaccination policies. Moeove, policies that combine vaccination in Paaguay and stamping out elsewhee wee shown to have the highest net benefits, given that such a combined policy led to slightly shote outbeaks and was less costly than pue stamping out. Pophylactic, peventative vaccination by neaby egions (i.e., vaccinating upon discovey of the disease in Paaguay) was the most effective stategy fom an epidemiological standpoint and was the best 2 The Southen Cone is defined as Agentina, Uuguay, Paaguay, and Southen Bazil.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 4 shot-tem policy, but esulted in educed benefits ove time, given the disuption vaccination policies have on accessing expot makets. One should obseve caution fom these esults, since the benefits fom any disease contol effot depend substantially on the ability of neighboing counties to effectively contol disease and may mitigate a successful stamping-out policy (Rich, et al., fothcoming). These esults thus suggest a coodinated multinational appoach towad disease contol. Moe impotantly, this pape demonstates the utility of an integated epidemiological-economic model to facilitate the spatial analysis of disease contol by measuing secto and welfae impacts while captuing maket and disease dynamics in a moe pecise way than most appoaches allow. An Oveview of FMD FMD is a vesicula disease affecting cloven-hoofed animals, such as cattle, pigs, sheep, goats, dee, and buffalo. infected with FMD develop bliste-like lesions on the mouth and foot. FMD is geneally not fatal in livestock, though motality in animals less than one yea of age is significantly highe; in swine, fo example, motality ates have been estimated at 80 pecent fo young animals less than twenty pounds (McCauley, 1979). In addition, pegnant livestock infected with FMD ae at significantly geate isk of spontaneous abotion. The main impact of FMD on infected livestock is educed poductivity. Infected animals often lose weight duing the couse of infection, consequently esulting in geate costs in feed and shelte. Infected daiy cattle geneally poduce less milk duing the infectious peiod. In most cases, animals ecove fom FMD without any pemanent ill-effects, though this is fa fom univesal (McCauley, 1979). The economic significance of an FMD outbeak is much geate than these poductivity effects might suggest because of the impact of the disease on maket access in intenational beef makets. Given the apid spead and high containment costs associated with FMD, counties that ae FMD-fee (as designated by the Intenational Office of Epizooties, o OIE) estict impots of meat fom counties that ae not FMD-fee, with tade limited to cetain types of meat (e.g., pocessed meat). Sanitay estictions on tade thus ceate a segmented maket in which fesh
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 5 meat expots fom counties that ae FMD-fee sell at a pice pemium (between 10-50 pecent) ove poducts that do not have this designation (Ekboi et al., 2002). 3 Moeove, cetain highvalue intenational makets, such as Japan and Koea, make a futhe distinction in commece between FMD-fee counties in which vaccination is pacticed and those that ae FMD-fee without vaccination, since it is difficult to ascetain the diffeence between meat fom an infected animal vesus one that has been vaccinated and geneated an immune esponse to the disease (Rich, 2004). This zeo-isk policy esticts meat impots in these makets fom all but FMDfee without vaccination souces. Tade estictions ceate poweful incentives to eliminate FMD in counties with expot potential, but the costs of doing so ae substantial. The counties of the Southen Cone have stuggled ove the past centuy to eadicate FMD fom thei cattle heds. Afte the egion successfully eadicated the vius in the mid-to-late 1990s and gained access to many new expot makets, FMD eappeaed in 2000-2001, esulting in significant expot losses. Many high-value makets emain inaccessible to expots fom much of this egion due to its disease status. 4 FMD contol stategies vay by county and context. A stamping out policy involves the slaughte of infected heds and heds in diect contact with infected heds, usually defined by those within a pe-set adius fom the infected heds. Ring vaccination is sometimes conducted in conjunction with a stamping out policy fo heds outside the contol zone to ceate a buffe aea to futhe contol the spead of disease. Movement contols ae also implemented. A pue stamping out appoach is to be followed in the case of an outbeak in the United States, though vaccination can be adopted unde cetain conditions (Ekboi, 1999). In counties whee FMD is endemic, vaccination is the pimay contol stategy, with contact slaughte and additional ing vaccination used to contol specific outbeaks. In Southen Afica, whee FMD is lagely spead by wildlife, FMD contol zones have been established in which the contol zone is suounded by two electified fences with a 1-km buffe aea. This stategy has been elatively successful in Southen Afica until ecently, when FMD outbeaks beached the FMD contol zones in 3 In addition, counties that ae FMD-fee have moe flexibility in maketing cetain types of cuts to divese makets. 4 Neithe Agentina no Paaguay is ecognized by the OIE as being FMD-fee with vaccination, due to outbeaks in 2003 and 2002, espectively. Uuguay, on the othe hand, has been ecognized by the OIE as FMD-fee with vaccination since May 2003 and has been able to maket many (but not all) types of beef expots to most majo makets except Japan, Koea, and Mexico.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 6 Zimbabwe and South Afica. In the Southen Cone, vaccination was employed to eadicate the disease in 1990s, afte which time stamping out was to be employed to teat isolated outbeaks. Because the massive scope of the 2001 outbeaks in Agentina and Uuguay pecluded stamping ou mass vaccination of cattle heds was used. 5 Applications of economics in animal disease models Standad models of animal disease typically use patial budgeting foms of benefit-cost analysis (BCA) in conjunction with epidemiological models of disease spead to assess the costs and benefits of altenative stategies (Mille, et al., 1996; Hos 1998; Nielen et al., 1999; Pey et al., 1999; Disney et al., 2001; Bates, 2002; Randolph et al., 2002). These models ae paticulaly useful at the hed and fam level and have the additional advantage of being tanspaent and easy-to-use (Rich, et al., 2004). Howeve, they ae unable to captue pice o welfae effects, linkages between sectos, and adjustment pocesses that can occu as a esult of an outbeak (cf. Beentsen, et al., 1992). In esponse to the limitations of benefit-cost models, seveal methodological appoaches have been used in moe ecent disease contol models; a thoough eview can be found in Rich, et al. (2004). Seveal studies have used input-output (I-O) models (o social accounting matices) to deive sectoal multiplies, which measue the economy-wide impact of a final demand shock in the livestock secto caused by a disease outbeak (Gane and Lack, 1995; Caskie et al., 1999; Ekboi, 1999; Mahul and Duand, 2000). Typically, multiplies ae computed fo labo makets, households, and livestock and elated sectos. An epidemiological model is usually used to calibate the size of the shock, which is used with the multiplies to compute the total impact of vaious disease stategies. While I-O models ae intuitively appealing, they suffe fom two main dawbacks that have been geneally ovelooked in pevious analyses. Fis input-output models ae fundamentally demand-diven models that assume supply is pefectly elastic. Howeve, long poduction cycles, paticulaly fo cattle, make this assumption poblematic; indeed, livestock supply may be pedetemined (Eales and Unneveh, 1993). Moeove, the 5 While sheep and pigs ae susceptible to FMD, vaccination pogams in the Southen Cone have only been pescibed fo cattle.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 7 natue of most animal diseases epesents both a supply shock and a demand shock. (BSE is an impotant exception). As a esul past studies likely ovestate the impact of disease outbeaks (Sadoulet and de Janvy, 1995). Secondly, pevious studies have failed to calculate the net impact of a disease, since an outbeak can geneate employment and income in othe sectos (e.g., govenment) depending on the mitigation stategy that can patially balance the negative impact of the disease. Rich (2003) uses an empiical model of FMD contol in Zimbabwe that incopoates supply constaints in the livestock secto to illustate the magnitude of these two effects. Aside fom I-O models, computable geneal equilibium (CGE) models have been used occasionally to model animal disease issues, but only aely (Pey et al., 2003) in concet with epidemiological models. Recent analyses of animal disease in Ieland (O Toole et al., 2002) and England (Blake et al., 2002) have teated disease-elated shocks to the economy as an exogenous shock, athe than one calibated fom a fomal disease spead model, fo instance. While CGE models have meit in thei ability to model economy-wide phenomenon, an agicultue-based shock such as an animal disease outbeak equies a detailed, agicultue-based social accounting matix to pefom an appopiate analysis. Patial equilibium models have also not been used with geat fequency in animal disease analysis. Amosson et al. (1979) used a patial equilibium model to evaluate the benefits of bucellosis contol. Beentsen, et al. (1992) used a single-secto patial equilibium model to deive welfae impacts fom altenative disease contol stategies of FMD in the Nethelands. Multisectoal models have only been used ecently in animal health applications. Mangen, et al. (2004) used a vetically-integated model of the hog industy in the Nethelands to analyze of Classical Swine Feve in the Nethelands; elated input and output makets wee not used in thei model, howeve. Schoenbaum and Disney (2003) used the USMP model oiginally designed by USDA-ERS in the mid-1980s to compute welfae effects of altenative FMD contol scenaios fo the United States. In the next section, the stuctue of DISCOSEM is povided to illustate the specification of an integated epidemiological-economic model of animal disease contol. The economic component of the model esolves many of the shotcomings of past economic analyses by using
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 8 a patial-equilibium, multi-maket model of animal disease contol that is both dynamic and spatial. Stuctue of DISCOSEM Epidemiological model 6 The disease contol potion of DISCOSEM is gounded in standad veteinay epidemiological models used in the analysis of FMD. Specifically, a state-tansition model is used (Daley and Gani, 1989). Biefly, a state-tansition model is a system of diffeential equations (o fistdiffeence equations if discete time is used) that epesents the tansition of animals between diffeent stages of disease. The most pevalent type of state-tansition model used fo FMD is an S-I-R model, which has thee states (Susceptible, Infected, and Removed). move between the Susceptible to Infected and Infected to Removed states based on tansition ates that ae eithe assumed fom past studies o calibated fom past outbeaks (Mille, 1979; Beentsen, et al., 1992; Gane and Lack, 1995; Ekboi, 1999; Duand and Mahul, 2000; Schoenbaum and Disney, 2003). In discete time, a state-tansition model is deived fom a Makov Chain and uses tansition pobabilities to chaacteize the movement of animals between states (Gane and Lack, 1995). The S-I-R system is usually closed such that S + I + R = N, whee N is the total numbe of animals in the system. Equation (1) chaacteizes the simple S-I-R system; the paametes β and α epesent the tansition ates fom Susceptible to Infected and Infected to Removed, espectively: ds = βsi dt di = αi + βsi dt dr = αi dt (1) Models of FMD contol have often modified the S-I-R system of (1) to incopoate additional states (Ekboi, 1999; Duand and Mahul, 2000; Schoenbaum and Disney, 2003). Fo instance, the Removed state can be patitioned into the states Immune (to incopoate vaccine esponse) 6 Moe detail on the epidemiological model can be found in Chapte 3 of Rich (2005).
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 9 and Dead (to eflect animals that have been culled o died fom the disease), while the Infected state can be sub-divided to eflect peiods of disease latency o exposue. In conventional models, the study aea is assumed to be adial, with the adius pedetemined based on the chaacteistics of the disease contol pogam (most studies assume a 10-30 km adius). Some models assume two adial aeas, with the oute aea epesenting a buffe vaccination zone. All animals (o heds 7 ) in the cicle ae assumed to be susceptible, with one animal (o hed) assumed to be infected at time t = 1. The S-I-R model is then un based on the assumed o estimated tansition ates. Diffeent mitigation stategies (vaccination, stamping out) ae simulated based on changes in the magnitude of the tansition ates and sequencing of othe exogenous paametes. DISCOSEM is based on a modification of the FMD model of Duand and Mahul (2000) (heeafte, DM). Unlike most conventional state-tansition models of disease contol, the DM model incopoates the spatial spead of disease within a paticula egion. In the DM model, two adial aeas ae assumed: an inne-ing with a adius of 10-km and an oute ing with adius 5-km. in the inne ing ae designated as in the state Susceptible-Exposed (S), while animals in the oute ing ae in the state Susceptible-Not Exposed (SN). In each time peiod, each ing is assumed to gow by 1 km, with some animals fom the SN state moving to the state S, while some animals fom outside the ing move into the SN state. Thus, unlike a conventional S-I-R model, the DM model is not closed. The DM model also futhe subdivides the Infected state into Incubating (o latent), Invasion, and Clinical states and the Removed state into Immune and Dead states. While the DM model is useful in modeling spatial spead within a egion, it does not explicitly handle inte-egional spead of disease. Howeve, a simple modification to the S-I-R famewok can be used that incopoates spatial spead. Rushton and Maunte (1954) fist modified the S-I- R model to examine mixing between heteogeneous populations by simply adjusting the βsi tem such that it incopoates the movement of infecteds fom within one s own egion (say, 7 In some models, including DISCOSEM, the unit of measuement is the hed athe than the animal, such that if one hed is infected, it is assumed that all animals in that hed ae also infected.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 10 egion i) and fom egions j = 1... k, i j, to egion i. A simila famewok was used by Howe et al. (2003) in a model of Bovine Malignant Catahal Feve in Tanzania. ds dt i = β S I ii di i = α i I dt dri = α i I i dt i i i S + β S I ii i k i j= 1 j i i β I ji + S j k i j= 1 j i β I ji j (2) The epidemiological potion of DISCOSEM is pogammed in STELLA (vesion 7.0.3) and its schematic is pesented in Figue 1. While complicated on fist glance, the diagam illustates the stuctue of the model. The paametes in the dak cicles epesent the stocks of animals in each of the seven states of the model (Susceptible-Exposed, Susceptible-Not Exposed, Incubation, Invasion, Clinical, Immune, and Dead). Paametes in the ectangula shapes ae the flows of animals between states. As DISCOSEM is a discete-time model (using one-week time steps 8 ), tansition pobabilities ae used to chaacteize movements between states; in the diagam, these pobabilities ae highlighted by the pentagonal shapes. The paamete p si (tansition pobability fom susceptible-exposed to incubation) paamete in the model is analogous to the β paamete in the continuous time model and is a function of the dissemination ates (DRs) fom each egion (10 in total: 8 in Agentina, Uuguay, and Paaguay). Fo egion i, the pobability of moving fom susceptible-exposed to incubation is a function of intenal spead within egion i and impots of animals fom each egion to egion i. Note that only adjacent egions ae assumed to contibute to the spead of disease fo a given egion (some links will thus be zeo). Moeove, it is assumed that tade between cetain egions is stochastic and may o may not occu in each peiod. Tems that ae highlighted with octagonal shapes epesent paametes that ae eithe specific to the disease contol stategy modeled (alpha, delta) o ae 8 This is contast to Duand and Mahul (2000) who use time steps of 0.5 weeks.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 11 tigges to tun on o off cetain paametes depending on the evolution of disease. Paametes highlighted with light oval shapes chaacteize the movement of the adial aea ove time. 9 Economic model Oveview The economic component of DISCOSEM is based on spatial equilibium models developed in the Makets and Stuctual Studies Division at the Intenational Food Policy Reseach Institute (IFPRI) in the late 1990s (Goletti and Mino 1998). DISCOSEM employs a solution technique known as mixed complementay pogamming (MCP) that is commonplace in the solution of CGE models. This is in contast to the majoity of spatial equilibium models which use quadatic pogamming o pice endogenous modeling techniques (Takayama and Judge, 1971; McCal and Speen, 1980) to deive optimal pices and movements of tade acoss egions. Quadatic pogamming models involve the maximization of poduce and consume suplus subject to flows of poducts acoss egions. In an MCP model, the equations that specify the model ae essentially the fist-ode, Kuhn-Tucke conditions of the quadatic pogamming model (Ruthefod, 1995). This yields a system of n equations and n unknowns, in which a subset of the unknowns ae the coesponding shadow pices (Lagange multiplies) fom the maximization poblem. Each inequality constaint is affiliated with a complementay vaiable (i.e., its shadow pice). If the inequality constain f(x) 0 is binding, an additional equation (λ > 0) must ente the system to ensue that the complementay slackness condition, λ f(x) = 0 holds. Unlike a quadatic pogamming model, MCP models have no objective function as they ae a squae n by n system. Howeve, the complementay vaiables must be associated with the elevant equations in GAMS in the model statement to ensue solution. An advantage of the MCP appoach ove quadatic pogamming models is in the flexibility the fome appoach povides the analyst. In a quadatic pogamming model, supply and demand cuves ae necessaily linea in ode to peseve the integability of the objective function. As a 9 Paametes that ae epeated in the diagam ae efeed to in STELLA as ghosts, which ae simply used to ease exposition by educing the numbe of aows epesented in the diagam.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 12 consequence, non-linea taxes (e.g., ad valoem taiffs) and non-linea demand systems (e.g., Rottedam o AIDS) cannot be used in a quadatic pogamming model. By contas an MCP model can utilize well-behaved, non-linea functional foms in both supply and demand equations, thus allowing fo the use of complex functional foms and systems. The economic component concentates on modeling phenomenon in the agicultual side of the economy. A total of six economic sectos ae modeled: cattle, beef, pok, lamb, con, and soybeans; cattle inventoies ae also included. 10 In the beef secto, quality components that diffeentiate beef cuts ae used to bette epesent the impact of FMD on expot makets. This entails sepaating the beef maket into a high-quality component and a low-quality component. High-quality cuts ae those that ae mainly taded on wold makets as chilled o fozen cuts, while low-quality cuts ae mainly consumed domestically. A five-yea time hoizon is used in the model eflecting the adjustment pocesses in tems of investment and animal inventoies esulting fom an outbeak. The model is solved ecusively, in which changes to animal inventoies, population, and pe capita income dive the data geneating pocess fo each peiod (Day and Cigno, 1978). The dynamic, long un natue of the model thus distinguishes it fom past models of FMD contol. Space is incopoated in DISCOSEM though the modeling of tade flows between thee egions in Agentina (Patagonia and Cuyo, Pampas, and the Noth of Agentina), Uuguay, and Paaguay; the thee egions of Agentina in the economic model ae aggegates of the eight epidemiological study egions. The inteactions between egions, in tems of egional tade, ae modeled explicitly in DISCOSEM. The advantage of this is to captue animal movements and egional income, as well as to model the diffeential effects of an outbeak on a egional basis. This is impotant in the context of the Southen Cone, paticulaly Agentina, given the specialization of egions in cetain types of poduction (beeding, fattening, slaughte). Unlike CGE models, non-agicultual sectos ae not explicitly modeled no ae capital, employmen o foeign exchange makets. The economic effects geneated by this model thus exclude many possible economic linkages and should be viewed as a fist-ound appoximation 10 Daiy poducts ae cuently excluded fom the model, despite the poductivity effects an FMD outbeak can have on this secto.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 13 to the tue impacts of any simulated shocks. Nonetheless, the choice of a patial equilibium model ove othe multisectoal appoaches (input-outpu CGE) was made fo a numbe of easons. Fis a patial equilibium model allows fo geate flexibility in modeling phenomenon in the agicultual secto than eithe I-O o CGE methods, paticulaly in a multi-egion, multicounty famewok. The Agentina I-O table, fo example, is a 73-secto model with sepaate sectos fo agicultue, livestock, meat poduction, and daiy poduction. Howeve, the Agentina I-O model is a national model, thus pecluding the staightfowad inclusion of egional impacts. Moeove, the level of detail in the Agentina I-O table is geate than that which exists fo Uuguay; it is unknown whethe a cuent input-output table exists fo Paaguay. Moe impotantly, the lack of commonality among sectos in diffeent I-O tables would make a detailed multi-egional I-O o CGE analysis poblematic. Secondly, the deficiencies of a patial equilibium model vis-à-vis an I-O o CGE model in the context of animal disease ae unlikely to have seious consequences in the cuent context. While patial equilibium models do not have the analytical powe to examine changes in employment and non-agicultual sectos (Rich, Mille, and Winte-Nelson, 2004), these issues ae less impotant in South Ameica than the detailed secto-level impacts povided in a patial equilibium model, given that the impact of an outbeak would be felt pimaily among livestock poduces and pocessos. National employment impacts fom an FMD outbeak would likely be modest and tempoay, and any national decline in employment in livestock poduction would be offset by a coesponding incease in govenment spending to combat the outbeak. Any effects on capital and foeign exchange makets would also be shot-lived. Non-agicultual impacts could be measued with an I-O table, using shocks fom the agicultual secto that wee obtained fom the multimaket analysis. Economic model specification The economic model is compised of five blocks of equations: supply, demand, income, pices, and tade. The fist and second block of equations denote the supply and demand elationships fo mea livestock, and feedgains. The model is set up as a vetically integated system using the equation specification of Jeong et al. (2003), with the exception that the fed cattle maket is not modeled due to data limitations. Also, all supply and demand equations ae modeled as
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 14 double-log, constant elasticity functions. A vetically integated system has been used in othe patial equilibium fomulations applied to animal disease contol (Mangen, Buell, and Mouits, 2004). Howeve, DISCOSEM is unique in this sense by modeling all majo meat sectos in addition to elated feed makets. A diagam of the patial-equilibium multimaket famewok of the model is pesented in figue 2. The input makets in the model include con and soybeans, which ae used fo feed. Othe inputs, such as hay and pastue which ae impotant in the poduction of cattle, ae not cuently modeled due to lack of data. Likewise, land, fetilize, and labo makets ae also not included as input makets. The equations in (3) specify con and soybean supplies (S) at time t in egion as functions of thei own poduce pice (pp). Cop demand (D, equations in (4)) is modeled as a function of its own consume pice (pc), the consume pice of substitute feeds, the poduce pice of pok (pp pok ), and, in the case of con, the poduce pice of slaughte livestock (pp sl ): S S D D con soy con soy S S con soy D D soy ( pp ( pp con con soy ( pc ( pc ) con ) con, pc, pc soy soy pp pp pok pok ), pp sl ) (3) (4) The thee equations fo animal inventoies and slaughte cattle makets specify the dynamics of investment and consumption behavio in live animal makets; this eflects the view that cattle (and othe livestock) ae both consumption and investment goods as chaacteized by Javis (1974, 1986). Following Jeong et al. (2003), cattle inventoies (INV) ae modeled in equation (5) as a function of the one-peiod lagged poduce pice of slaughte cattle and the one-peiod lagged supply of (exogenous) beef calves. The supply of slaughte cattle (S sl ) will depend on cuent cattle inventoy, lagged supply of slaughte cattle (eflecting a patial adjustment pocess), the poduce pice fo stees, and the consume pice of con. 11 While con is not a majo input to cattle poduction, it is included to captue the inceasing use of feedlot poduction, paticulaly in Agentina. This inclusion of animal inventoies and live animal makets enables DISCOSEM to examine the ole of long-tem investment and thus extends past multimaket 11 The Jeong et al. (2003) specification includes expectations about pices; this is not modeled in this famewok.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 15 models (Baveman et al., 1987; Goletti and Rich, 1998) that have chaacteized livestock makets in a much simple manne. Demand fo slaughte cattle (D sl ) is a function of cuent slaughte pices at the consume level, lagged demand, and the poduce pice fo high-quality (HQ) and low-quality (LQ) beef. sl INV INV (pp t 1,,Q calves t 1, ) (5) sl S S sl sl (S t 1,, pp sl, pc con,inv ) (6) D sl sl sl sl HQ LQ D D, pc, pp, pp ) (7) ( t 1, Fo pok and lamb, meat supply (equations in 8) is modeled on the basis of lagged supply and the own poduce pice of meat. The pok maket is also a function of feed pices (con and soybeans). In the case of beef, high-quality and low-quality beef ae assumed to be fixed popotions (λ and 1-λ, espectively) of cattle demand conveted into etail equivalent based on slaughte and technical convesion factos; thus cacass makets ae not diectly modeled. This is clealy a compomise, since beef supply should ideally be a function of own and input pices. Howeve, modeling limitations cuently pevent econciling a flexible beef supply specification with the estiction that total beef supply be equal to livestock demand conveted to etail beef equivalent. 12 As a esul the pice effects esulting fom a shock to meat makets will ovestate the tue effects. Meat demand (D m in equation 9) is modeled as a function of own consume pice, the consume pice of substitutes, and income pe capita (YPC). S S S pok lamb HQbeef S S pok lamb ( S ( S sl pok t 1, lamb t 1,, pp, pp LQbeef pok lamb, pc con λd, S (1 λ) D ), pc sl soy ) (8) { pok, lamb, HQbeef, LQbeef } m m HQ LQ pok lamb D D ( pc, pc, pc, pc, YPC ), m (9) 12 The issue in question is as follows. Ideally, thee must be balance between livestock demand (conveted to etail equivalent) and total etail beef supply, given that etail beef supply deives diectly fom livestock demand. The poblem is that any such explicit estiction conflicts with the MCP fomulation of the model (i.e., the system is no longe squae). The solution is to define some appopiate complementay vaiable to match this estiction, but this definition has not been adequately tested to date. As a esul a fixed, Leontief appoach is utilized, with futhe eseach aimed at esolving this poblem.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 16 Thee inequalities detemine the movement of pices in the model. In the domestic make the poduce pice (pp) of a commodity g in egion plus tanspotation costs (TC) and commecial magins (MARGD, eflecting makups fom wholesale to etail) must be at least as lage as the consume pice (pc) in egion ; if the constaint is binding, thee will be tade between egion and egion. Likewise, the consume pice (pc) must be less than o equal to the impot pice (pm) plus tanspotation costs and impot magins (MARGM, eflecting the makup fom the pot to etail), while the expot pice (px) should be less than o equal to the poduce pice (pp) plus tanspot costs and expot magins (MARGX, eflecting the makup fom the fam to the pot). If eithe equation is binding, thee will be enty of (espectively) impots and expots into the system, viz.: pp g g + TC, + MARGD g g pc (10) pc g pm g t + TC g g + MARGM (11) px g t pp g + TC g g + MARGX (12) Inflows and outflows of commodities acoss egions ae egulated by equations 13 and 14. Fis total demand (D) fom egion must not exceed total impots (I) to egion fom the est-of-thewold and the sum of tade (TQ) fom all othe egions () to egion. Second, total supply must be at least as lage as total expots (X) fom egion to all othe egions (TQ) plus expots aboad: g g g TQ t, + I D, (13) g g g S TQ, + X (14) The final block of the model is the income block, which, fo each egion, is simply the sum of agicultual income pe capita plus exogenous non-fam income pe capita (NFYPC). Fam income is defined as the net sum of evenue fo each poduct: YPC = NFYPC + ( m pp m m S + pp sl sl S + pp c c c S m { pok,lamb,hqbeef,lqbeef };c con,soybeans { } pc sl D sl t c pc c c D )/POP (15)
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 17 As mentioned ealie, DISCOSEM is solved in GAMS using the MCP solve; thus no objective function is equied, povided the model is specified with an equal numbe of equations and unknowns and popely defined complementaity conditions. Dynamics ae simulated by solving the model ecusively (Day and Cigno, 1978). Data A majo challenge in calibating DISCOSEM is that some of the egional data fo Agentina and national data fo Paaguay ae not available. National data wee used to constuct egional databases in many instances. Baseline data on poduction fo livestock and cops wee fom the Ministies of Agicultue of Agentina, Uuguay, and Paaguay. These data included infomation on animal demogaphics, the numbe of animals slaughteed, and, fo Agentina and Uuguay, statistics on the aveage cacass weight fo slaughteed animals. FAO data wee used in some cases to compute aveage cacass weight and slaughte (Paaguay) and to obtain infomation on pok and lamb poduction (Uuguay and Paaguay). Animal demogaphic data was used to compute the aveage numbe of heds and pe-km hed density in each egion fo the epidemiological model. Data on epidemiological paametes wee deived fom a numbe of souces. Detailed, unpublished daily data on the 2000-2001 FMD outbeak in Agentina was obtained fom the National Animal Health and Agifood Quality Sevice (SENASA) of Agentina and used to calculate the inta-egional dissemination ates of disease used in the model. Weekly outbeak infomation was available fo Uuguay, while no data was known fo Paaguay and thus data fom Nothen Agentina was used to poxy disease spead in Paaguay. Two sets of intaegional DRs wee used: one ate to simulate the spead of disease befoe the fist few clinical cases and a lowe ate that eflects the impact of disease contol effots. While the outbeak data povides infomation on the spead of the disease within a egion, it does not assess inteegional tade. Inte-egional DRs ae thus calculated by taking the total egional DR as calculated fom the data and subdividing it into inta-egional and inte-egional components on
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 18 the basis of appoximate animal movements between egions. 13 As with inta-egional DRs, two sets of inte-egional DRs wee used; afte discovey of disease in a egion, they ae usually set to zeo, but in majo egions (e.g., the Pampas), they ae assumed to be small and stochastic. Both inta- and inte-egional DRs ae assumed to be stochastic and unifomly distibuted on a egion-specific ange. Because tade between egions is stochastic, any given simulation of the model can give esults that ae small (if no tade occus) o much lage. The emaining tansition pobabilities ae based on those used by Duand and Mahul (2000). The aeas used in the model ae lage than those used in the DM model (30-km inne ing, 10-km oute ing) to eflect the extensive livestock systems of the Southen Cone. The size of the inne ing also eflects the size of the contol aea of the 2000 Atigas outbeak in Uuguay. Fo the economic potion of the model, cetain statistics, such as pices, wee somewhat poblematic. In the case of Agentina, monthly time seies data ae available fo the slaughte pice of animals by age categoy (calves, stees, heifes, cows, and bulls) fo the Linies make which is the lagest auction yad in Agentina. Howeve, egional data on slaughte pices ae not consistently available. Fo instance, the website fo SAGPyA (Ministy of Agicultue, Livestock, and Fisheies) in Agentina contains spoadic infomation on slaughte and fed cattle pices fo cetain disticts on a monthly basis, but the fequency of updates is inconsistent. Moeove, time seies data of this natue ae not easily available. Peliminay analysis of data collected fo 2003 suggests that the slaughte pices of outlying egions (Patagonia/Cuyo and the Noth) ae slightly lowe than the Linies pice. Fo the puposes of the model, it was assumed that the Patagonia/Cuyo pice is 10 pecent lowe than the Linies pice (used to poxy the Pampas pice), while the pice in the Noth was assumed to be 5 pecent lowe. Slaughte pices wee available fo Uuguay, but not Paaguay. Anecdotal evidence suggests that the live animal pices in Paaguay ae significantly lowe than those in Agentina; in 2003, a pess aticle epoted that pices in Paaguay wee 40 pecent lowe than those in Agentina. Theefoe, the live animal pice fo Paaguay was assumed to be 40 pecent lowe than the pice in the Noth. 13 In some instances, the inte-egional DRs ae lage than inte-egional tade would suggest. One should note that disease spead can occu fom othe pathways, such as coss-egional taffic, aibone spead, and wildlife, which would justify a slightly highe spead ate.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 19 Data on demand and etail pices wee of vaying quality. Fo Agentina, detailed infomation on etail pices by type of beef cut by egion was ecoveed fom the 1996/97 Household Suvey conducted by INDEC. Using this data and insights fom expet analysis, cuts of beef wee aggegated into high and low quality components and aveage pices computed fo each egion. These pices wee conveted to 1999 pices using data on consume pice inflation by food poduct. Fo Uuguay, etail pices fo fou diffeent cuts wee available fom the Ministy of Agicultue, Livestock, and Fisheies, and categoized into high and low components, espectively. Data on pe capita consumption fo Agentina was calculated fom the INDEC household suvey. Pe capita consumption fo Uuguay and Paaguay was deived as the esidual fom food availability (poduction plus impots less expots). Elasticities ae to be econometically estimated based on infomation fom the household suvey and time-seies data on livestock and feed cops. Fo the time being, elasticities fo livestock and meat poducts ae based on Jeong et al. (2003). 14 Simulation analysis The stuctue of DISCOSEM is amenable to the analysis of altenative disease mitigation stategies. In the liteatue, conventional analyses have eithe used an epidemiological model to calibate the disease shock (Beentsen, et al., 1992; Gane and Lack, 1995; Ekboi, 1999; Schoenbaum and Disney, 2003) o have simply assumed exogenous shocks that would coespond to a disease outbeak (O Toole et al., 2002); DISCOSEM epesents the fist vaiety of model, given its integative natue. The simulation analysis poceeds as follows. At time t = 1, an outbeak is seeded in Paaguay though the intoduction of one hed in the state Incubation. The location of the outbeak in Paaguay is not paticulaly impotant. In the next peiod, t = 2, the disease will spead in Paaguay though the expansion of the two ings as descibed ealie; the definition of the tansition pobabilities ae such that discovey of a clinical case does not occu until t = 4. 14 Econometic estimates fo the elasticities of the model ae pending and will be incopoated in the final esults (Rich, 2005). Javis (1974, 1986) has a numbe of detailed models of the livestock secto in Agentina, but does not povide sample means to compute elasticities with his linea model, thus pecluding the use of his esults.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 20 Because the possibility of tade between Paaguay and its two adjoining egions is stochastic, it is possible that at t = 2 an outbeak will commence in one o both neighboing egions. In the model, it is assumed that thee is a five pecent chance in any peiod that such tade happens. If tade occus, the disease will spead within these egions and then (in subsequent peiods) spead southwads (as well as westwads and eastwads) to adjoining egions. In figue 3, a hypothetical illustation of disease spead with the epidemiological component is povided. In the top fame (t=1), the disease is located in a adial aea in Paaguay. In the next peiod (t=2, o the bottom fame), if thee is tade, thee will be the spead of disease fom Paaguay to the two egions to the south of Paaguay. In addition, the adial aea within Paaguay will gow in the next peiod. Note that the spead of disease fom Paaguay to these adjoining egions could also occu with some delay o neve occu and thus be isolated to Paaguay. Once the disease is diagnosed in a egion, policymakes can apply one of thee mitigation stategies. Fis one could apply a stamping out policy, which in this model efes to slaughteing all obviously sick animals (i.e., those in the Clinical state) and susceptible animals in potential contact with sick heds. 15 Secondly, a vaccination policy could be applied in which obviously sick animals ae culled and all susceptible animals ae vaccinated. A thid policy is what is temed peventative vaccination, which epesents an abstaction of a pue vaccination policy. In this stategy, once the disease is diagnosed in Paaguay, all egions pophylactically vaccinate thei heds beginning in vulneable aeas (e.g., auction yads that eceive high volumes of tade) and inceasing the scope of vaccination in a adial fashion fom these aeas. Six scenaios wee un with the epidemiological model based on combinations of the above stategies and ae summaized in table 1. Because of the stochastic natue of the dissemination ates, Monte Calo simulations of the model wee conducted by compiling the STELLA model equations into Bekeley MADONNA (vesion 8 Beta 10) and unning 1000 batch uns fo each scenaio. Fom these simulations, the aveage length of outbeak, numbe of animals culled, and numbe of animals vaccinated wee obtained. In addition, a lage outbeak of each scenaio was simulated by using the mean plus standad deviation of these esults fo a total of twelve simulations oveall. 15 In the DM model, this scenaio is efeed to as the SODC stategy (Stamping Out Diect Contact heds).
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 21 The economic component is un ove a five-yea peiod and it is assumed that an outbeak occus in yea 2. The epidemiological model povides two majo pieces of input into the economic module of DISCOSEM. Fis the aveage numbe of animals culled is used to define the shock to inventoies duing the outbeak yea; in geneal, this will be small (less than 1 pecent). Second, the aveage length of outbeak is used to calibate the amount that expots ae disupted in the outbeak yea. It is assumed that yea 2 expots ae esticted to the pevious yea s expots times a deflation facto, defined as 1-((# of outbeak weeks+16)/52). Following Ekboi (1999), it is assumed that it takes an additional 16 weeks afte the outbeak has been contained fo intenational tade to esume. ounded to the next highest (whole) week. It is also assumed that the numbe of outbeak weeks is The adjustment of wold pices depends on the mitigation stategy pusued. In all cases, thee will be a fall in wold pices eceived in the outbeak yea (yea 2) as cetain high-value makets will be closed. A 25 pecent dop in the expot pice is assumed fo yea 2. In the following yea (yea 3), moe high-value makets ae assumed to open and expot pices ae assumed to ecove to 90 pecent of thei fome value. In a vaccination (o peventative vaccination) stategy, FMD-fee without vaccination makets (and some FMD-fee with vaccination makets) will be closed fo at least two yeas afte the last vaccination; this will not be the case in a stamping out policy. Coespondingly, expot pices in a vaccination policy will emain at 90 pecent of the oiginal pice in yea 4 and only ecove in yea 5. In a stamping out policy, wold pices ae assumed to etun to nomal in yea 4. Costs fo each stategy ae detemined fom the epidemiological model and exogenous peanimal contol costs. Indemnity payments fo culled animals ae made on the basis of the slaughte value of the animal ($287, based on the aveage pice and live weight fo animals sold to the Linies auction yad in Agentina in 1999). 16 In addition, othe costs to animal slaughte (e.g., cleaning and disposal costs) ae assumed at $48 pe animal based on figues fom Schoenbaum and Disney (2003) fo a hypothetical U.S. outbeak. Vaccination and administation costs ae estimated at 90 cents pe animal pe yea (two vaccinations in the 16 This figue is likely high. In the Atigas outbeak in Uuguay, the aveage indemnity payment pe animal was $175. Howeve, the majoity of those animals wee sheep, which fetch a lowe value than beef cattle.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 22 outbeak yea at 45 cents each) (Rich, 2004). Heds ae assumed to be vaccinated in the outbeak yea only. Results Epidemiological Model The esults fom the epidemiological model ae summaized in table 2. Stamping out policies, not supisingly, lead to the lagest numbe of animals being culled. In the aveage outbeak scenaio, nealy two and a half times moe animals ae culled in aggegate unde stamping out than in the pue vaccination policy (table 2, (a)). At the same time, the duation of a stamping out policy vis-à-vis a pue vaccination policy is about one-half of a week shote on aveage. Cuiously, an exception to this is the povince of Ente Rios, which has a slightly longe outbeak unde stamping out than vaccination. Lage outbeaks incease the diffeence in the outbeak duation between stamping out and vaccination, with stamping out stategies lasting about 1.5 weeks less than vaccination stategies. Peventative vaccination stategies lead to makedly shote outbeaks than eithe stamping out o pue vaccination. With the exception of Paaguay, which is the epicente of the outbeak, outbeaks in the aveage scenaio have a duation of less than a week (table 2, (b)). Moeove, fou egions (Cuyo, Patagonia, Uuguay, and NOA) do not epot any outbeaks at all. In the lage scenaio, simila esults ae epoted. Fom an epidemiological standpoin a peventative vaccination policy is extemely effective in educing the incidence of disease. One should be cautious in intepeting these esults, howeve, since this type of contol stategy is an abstaction that assumes pefect tageting of sensitive aeas in neaby egions that might be infected fom an outbeak in Paaguay. At the same time, it highlights that apid mobilization of contol effots can have a significant impact on limiting the couse of disease. Combinations of these policies (i.e., distinguishing between stategies in Paaguay and the est of the Southen Cone) give esults that ae simila to the pue contol stategies. A VAC-SO policy in which Paaguay vaccinates while the est of the Southen Cone adopts the SO stategy leads to slightly shote outbeaks (with the exception of Paaguay) than a univesal stamping out
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 23 stategy. Fewe animals ae also culled. This implies that thee may be egional benefits to diffeential contol stategies acoss the Southen Cone and suggests that coodination of contol stategies acoss bodes could lead to geate success in educing the impact of an FMD outbeak. Economic Model The esults fom the economic model ae summaized in tables 3 though 5. In the shot-un, a peventative vaccination pogam is the optimal stategy fom the standpoint of the net evenue geneated fom such an appoach (table 3). In addition to being less costly than othe mitigations, a peventative vaccination policy causes fewe dislocations to expots, given that unde peventative vaccination an outbeak lasts, on aveage, about 2-3 weeks less than in othe contol stategies. As a esul the volume (and thus value) of expots is geate in the yea of the outbeak unde a peventative vaccination policy. In the long-un, the benefits to both types of vaccination policies ae educed, ostensibly due to geate disuptions to expots ove time vis-à-vis stamping out stategies (table 4). The ealie access to high-value expot makets engendeed by stamping out policies in late yeas (i.e., gaining access to high-value makets in yea 4 athe than yea 5) tanslates into a net pesent value ove a five-yea peiod that is $400-600 million highe than eithe pue o peventative vaccination policies (table 4). Inteestingly, the VAC-SO stategy in which Paaguay vaccinates and the est of the Southen Cone stamps out has the highest net pesent value -- in the aveage scenaio, the diffeence in net pesent value is oughly $15 million. Moeove, all egions, including Paaguay, benefit fom such a stategy (table 5). The intuition behind these esults is that a VAC-SO stategy is slightly shote in Uuguay (thus disupting expots less than SO) and slightly less costly (due to fewe animals being culled) and thus has lowe pice impacts in the outbeak yea than the SO policy. Since Paaguay is not an expote outside the Southen Cone, the effects of it vaccinating (and thus not having access to high-value makets fo an exta yea) ae muted. In addition, slightly highe pices unde VAC-SO in the est of the Southen Cone buoy pices fo poducts expoted by Paaguay within in the Southen Cone. By contas in the lage scenaio, while the VAC-SO stategy has a highe net pesent value than SO, the egional impacts of such a stategy ae mixed (table 5). In paticula, the Noth and
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 24 Uuguay ae bette off unde a SO contol policy. Unlike the aveage scenaio, the duation of an outbeak unde the lage scenaio in the VAC-SO stategy is simila to that of the SO stategy (due to the ounding of the outbeak weeks assumed in the deflation facto). Given the inteaction in egional tade in cattle and low-quality beef between Uuguay, the Noth, and the Pampas, the longe duation of the VAC-SO policy in the lage scenaio dampens pices and tade enough in these egions to slightly educe evenues in the outbeak yea. A numbe of caveats should be given to the peliminay analysis above. Fis expot pices ae assumed to emain constant in the scenaios, which is not ealistic. A small county assumption is also maintained, which may not be suitable fo the Southen Cone in meat makets. 17 simulations also conside pefect contol stategies in which counties of the Southen Cone emain FMD-fee without vaccination afte an outbeak. Howeve, such a scenaio is contingent on the disease status of neighboing and neaby counties (Bolivia, Peu, and Bazil). Indeed, ecent outbeaks that occued in Agentina and Paaguay in 2003 despite vaccination in these two counties demonstate the fagility of cuent inte-egional contol effots. The Conclusions The impacts of disease vay ove time and space. The peceding analysis illustates how animal disease can be examined in a way that captues both tempoal and spatial factos in an integative epidemiological-economic fashion. In the scenaios fo FMD contol pesented above, stamping-out policies wee shown to have a lage net pesent value ove a five-yea peiod than a vaccination-only stategy; howeve, such a policy might not be viewed as optimal in a shotun famewok. Multi-peiod analysis that focused on the epidemiological pogession of FMD but failed to captue economic behavio concening inventoies would also misepesent the evolution of costs and benefits ove time. The esults assume that both vaccination and stamping out could be implemented pefectly and that thee ae no spillove effects fom neighboing o neaby egions. Howeve, as 17 Indeed, it may be moe appopiate to model the Southen Cone as a lage county, since the expeience of Agentina and Uuguay afte the FMD outbeak of 2001 was a etun to pe-outbeak levels of expots the following yea, albeit at a much lowe expot value.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 25 demonstated in Rich, et al. (fothcoming), egional extenalities can play a significant ole in explaining the pesistence of FMD in South Ameica. The tendency of FMD to spead apidly ove space thus suggests that disease contol effots need to be caied out though a continentwide appoach athe than on a sub-egional basis. The esults of this analysis, while tentative, illustate a need fo sensitivity to the egional divesity of contol stategies. Optimal contol stategies will vay within the continent. To the extent that a given contol stategy is implemented fully and effectively, the divesity in appoaches need not undemine efficacy in any given sub-egion and may in fact enhance it. Spatial and dynamic policy analysis can play key oles in detemining the types of inteventions necessay in such an envionment. Refeences Amosson, S.H., Dietich, R.A., Talpaz, H., and Hopkin, J.A. (1979). Economic and Epidemiological Policy Implications of Altenative Bovine Bucellosis Pogams, Westen Jounal of Agicultual Economics, 6(1):43 56. Bates, T.W. (2002). Evaluation of stategies to contol tansmission of foot and mouth disease vius in a high-density livestock egion of Califonia by use of an epidemic simulation model. Unpublished PhD dissetation, Univesity of Califonia, Davis. Beentsen, P.B.M., Dijkhuizen, A.A., and Oskam, A.J. (1992). A dynamic model fo cost benefit analyses of foot-and-mouth disease contol stategies, Peventive Veteinay Medicine, 12:229 243. Blake, A., Sinclai, M.T., and Sugiyato, G. (2002). The Economy-Wide Effects of Foot and Mouth Disease in the UK Economy, In Thid Annual Confeence Euopean Tade Study Goup, Sep 2001. Bullock, D.S., Gacia, P., and Lee, Y.-K. (1996). Towads Poduce Welfae Measues in a Dynamic, Stochastic Famewok, Woking pape, Univesity of Illinois at Ubana-Champaign. Caskie, P. Davis, J., and Moss, J.E. (1999). The economic impact of BSE: a egional pespective, Applied Economics, 31(12):1623 1630. Daley, D.J. and Gani, J. (1989). Epidemic Modelling: An Intoduction (Cambidge, UK: Cambidge Univesity Pess). Day, R.H. and Cigno, A., eds. (1978). Modelling Economic Change: The Recusive Pogamming Appoach (Amstedam: Noth-Holland). Disney, T.W., Geen, J.W., Fosythe, K.W., Wiemes, J.F., and Webe, S. (2001). Benefit-cost analysis of animal identification fo disease pevention and contol, Revue Scientifique et Technique de l Office Intenational des Epizooties, 20(2):385 405.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 26 Duand, B. and Mahul, O. (2000). An extended state-tansition model fo foot-and-mouth disease epidemics in Fance, Peventive Veteinay Medicine, 47:121 139. Eales, J.S. and Unneveh, L.J. (1993). Simultaneity and Stuctual Change in U.S. Meat Demand, Ameican Jounal of Agicultual Economics, 75(2):259 268. Ekboi, J.M. (1999). Potential Impact of Foot-and-Mouth Disease in Califonia: The Role and Contibution of Animal Health Suveillance and Monitoing Sevices. Davis, CA: Agicultual Issues Cente, Univesity of Califonia, Davis. Ekboi, J., Javis, L.S., Sumne, D.A., Bevejillo, J.E., and Sutton, W.R. (2002). Changes in Foot and Mouth Disease Status and Evolving Wold Beef Makets. Agibusiness, 18 (2):213 229. Gane, M.G. and Lack, M.B. (1995). An evaluation of altenate contol stategies fo foot-and mouth disease in Austalia: a egional appoach, Peventive Veteinay Medicine, 23:9-32. Goletti, F. and Mino N. (1998). Expot Libealization and Household Welfae: The Case of Rice in Viet Nam, Ameican Jounal of Agicultual Economics, 80(4):738-749. Goletti, F. and Rich, K. (1998). Analysis of Policy Options Fo Income Gowth and Povety Alleviation, Repot pepaed fo the USAID poject on Stuctue and Conduct of Majo Agicultual Input and Output Makets and Response to Refoms by Rual Households in Madagasca, Washington, DC: Intenational Food Policy Reseach Institute. Hos H.S. (1998). Risk and economic consequences of contagious animal disease intoduction. (Wageningen and Leiden: Wageningen Agicultual Univesity and Backhuys Publishes). Howe, R.H., Boone, R.B., DeMatini, J.C., McCabe, J.T., and Coughenou, M.B. (2003). A Spatially Integated Disease Risk Assessment Model of Bovine Malignant Catahal Feve in Ngoongoo Consevation Aea, Tanzania, Woking pape, USDA-APHIS, Coloado State Univesity, and Univesity of Coloado. Javis, L.S. (1974). Cattle as Capital Goods and Ranches as Potfolio Manages: An Application to the Agentine Cattle Secto, Jounal of Political Economy, 82:489 520. Javis, L.S. (1986). Supply Response in the Cattle Industy: The Agentine Case. Giannini Foundation Special Repot. (Oakland, CA: Giannini Foundation of Agicultual Economics, Univesity of Califonia). Jeong, K.-S., Gacia, P., and Bullock, D.S. (2003). A statistical method of multimaket welfae analysis applied to Japanese beef policy libealization, Jounal of Policy Modeling, 25:237 256. Mahul, O. and Duand, B. (2000). Simulated economic consequences of foot-and-mouth disease epidemics and thei public contol in Fance. Peventive Veteinay Medicine, 47:23 38. Mangen, M.-J.J., Buell, A.M., and Mouits, M.C.M. (2004). Epidemiological and Economic Modelling of Classical Swine Feve: Application to the 1997/98 Dutch Epidemic, Agicultual Systems 81: 37-54. McCal, B.A. and Speen, T.H. (1980). Pice Endogenous Mathematical Pogamming as a Tool fo Secto Analysis, Ameican Jounal of Agicultual Economics, 62(1):87-102. McCauley, E.H. (1979). Estimation of the physical loss in animals infected with foot-and mouth disease. In E.H. McCauley, N.A. Aulaqi, J.C. New, J., W.B. Sundquis W.M. Mille, (eds.), Potential Economic Impact of Foot-and-Mouth Disease in the United States. St. Paul, MN: Univesity of Minnesota and APHIS-USDA. Mille, G.Y., Tsai, J.S., and Foste, D.L. (1996). Benefit-cost analysis of the national pseudoabies vius eadication pogam, Jounal of the Ameican Veteinay Medical Association, 208(2):208 213.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 27 Nielen, M., Jalvingh, A.W., Meuwissen, M.P.M., Hos S.H., and Dijkhuizen, A.A. (1999). Spatial and stochastic simulation to evaluate the impact of events and contol measues of the 1997-1998 classical swine feve epidemic in The Nethelands II: Compaison of contol stategies, Peventive Veteinay Medicine, 42:297 317. O Toole, R., Matthews, A., and Mulvey, M. (2002). Impact of the 2001 Foot and Mouth Outbeak on the Iish Economy, IMAGE poject woking pape, Tinity College, Dublin. Pey, B.D. et al. (1999). The economic impact of foot and mouth disease and its contol in South East Asia: a peliminay assessment with special efeence to Thailand, Revue Scientifique et Technique de l Office Intenational des Epizooties, 18(2):478 497. Pey, B.D., Randolph, T.F., Ashley, S., Chimedza, R., Foman, T., Moison, J., Poulton, C., Sibanda, L., Stevens, C., Tebele, N., and Yngstom, I. (2003). The impact and povety eduction implications of foot and mouth disease contol in southen Afica, with special efeence to Zimbabwe. Repot commissioned by the Depatment fo Intenational Development of the Govenment of the United Kingdom, Intenational Livestock Reseach Institute. Randolph, T.F., Pey, B.D., Benigno, C.C., Santos, I.J., Agbayani, A.L., Coleman, P., Webb, R., and Gleeson, L.J. (2002). The economic impact of foot and mouth disease contol and eadication in the Philippines, Revue Scientifique et Technique de l Office Intenational des Epizooties, 21(3):645 661. Rich, K.M. (2003). Supply constaints and mixed multiplies in input-output models of animal disease contol: Applications to Zimbabwe, Pape pesented at the 50th Annual Noth Ameican Meetings of the Regional Science Association Intenational, Philadelphia, PA, Nov. 22-23, 2003. Rich, K.M. (2004). Animal Diseases and the Cost of Compliance with Intenational Standads and Expot Makets: The Expeience of Foot-and-Mouth Disease in the Southen Cone, Unpublished epot pepaed fo the Wold Bank. Rich, K.M. (fothcoming, 2005). Spatial Models of Animal Disease Contol in South Ameica: The Case of Foot-and-Mouth Disease, Unpublished Ph.D. dissetation, Univesity of Illinois at Ubana-Champaign. Rich, K.M., Mille, G.Y. and Winte-Nelson, A. (2004). A Review of Economic Tools Relevant fo Consequence Assessment of Animal Disease Outbeaks. Submitted to Revue Scientifique et Technique de l Office Intenational des Epizooties. Rich, K.M., Winte-Nelson, A., and Bozovic, N. (fothcoming). Regionalization and foot-and-mouth disease contol in South Ameica: lessons fom spatial models of coodination and inteactions, Quately Review of Economics and Finance. Rushton, S. and Mautne, A.J. (1954). The deteministic model of a simple epidemic fo moe than one community, Biometika, 42:126 132, 1954. Ruthefod, T. (1995). Extension of GAMS fo complementaity poblems aising in applied economic analysis, Jounal of Economic Dynamics and Contol, 19:1299 1324. Sadoule E. and de Janvy, A. (1995). Quantitative Development Policy Analysis. Baltimoe, MD: Johns Hopkins Univesity Pess. Schoenbaum, M.A. and Disney, W.T. (2003). Modeling altenative mitigation stategies fo a hypothetical outbeak of foot-and-mouth disease in the United States, Peventive Veteinay Medicine, 58:25 52. Takayama, T. and Judge, G.G. (1971). Spatial and Tempoal Pice and Allocation Models. Amstedam: Noth-Holland Publishing Company.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 28 Table 1: Summay of Epidemiological Simulations used in the Model Disease Contol Stategy Desciption 1) Stamping Out (SO) All heds that ae obviously sick (in state Clinical ) ae slaughteed with pobability 0.7 o 0.8 depending on the stage of the outbeak. Also, heds in contact with sick heds ae slaughteed at ate of α = 2 heds/week in the initial peiod of disease, α = 5 heds/week in subsequent peiods. 2) Stamping Out in Paaguay, Vaccination in Rest of Southen Cone (SO-VAC) 3) Stamping Out in Paaguay, Peventative Vaccination in Rest of Southen Cone (SO-PREV) 4) Vaccination in Paaguay, Stamping Out in Rest of Southen Cone (VAC-SO) Same as (1) fo Paaguay. Fo est of Southen Cone, animals ae vaccinated upon enty of the disease in a given egion by change of pobability fom state Susceptible-Exposed o Susceptible-Not Exposed to Immune. This pobability is 0.7 fo Agentina, 0.9 fo Uuguay (initial pobability is zeo). Same as (1) fo Paaguay. Fo est of Southen Cone, animals ae vaccinated upon discovey of disease in Paaguay. Tansition pobabilities ae the same as (2). Fo Paaguay, tansition pobability fom state Susceptible- Exposed o Susceptible-Not Exposed to Immune is assumed to be 0.5. Stamping Out in est of Southen Cone follows desciption in (1). 5) Peventative Vaccination (PREV) Same as (3) fo est of Southen Cone. Paaguay natually follows a standad vaccination policy. 6) Total Vaccination (VAC) Tansition pobabilities follow (2) fo est of Southen Cone and (4) fo Paaguay.
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 29 Table 2: Summay of epidemiological simulations (a) Aveage Outbeaks Region SO SO-VAC SO-PREV Vaccinated Culled Duation (weeks) Vaccinated Culled Duation (weeks) Vaccinated Culled Duation (weeks) Pampas - 8,639 2.94 396,183 3,633 3.45 1,858,434 163 0.29 Cuyo - 1,931 2.15 96,522 1,159 2.90 217,414 - - Patagonia - 6,189 2.67 19,131 2,154 3.19 50,458 - - Uuguay - 3,206 2.09 948,198 1,723 2.56 2,815,593 - - Paaguay - 9,363 6.58-9,418 6.66-9,392 6.62 B. Aies - 33,985 3.24 638,404 11,931 3.92 3,073,369 436 0.49 NOA - 1,148 2.12 76,701 699 2.88 156,223 - - NEA W. - 3,575 2.77 350,323 1,999 3.61 1,515,527 151 0.51 NEA E. - 2,166 2.26 286,639 1,251 2.97 1,096,956 80 0.19 Ente Rios - 15,046 4.31 551,044 3,685 3.88 2,512,150 116 0.30 TOTAL - 85,248 3,363,143 37,652 13,296,124 10,338 Region VAC-SO PREV VAC Vaccinated Culled Duation (weeks) Vaccinated Culled Duation (weeks) Vaccinated Culled Duation (weeks) Pampas - 8,439 2.87 1,804,727 153 0.28 381,195 3,567 3.37 Cuyo - 1,885 2.08 210,670 - - 92,578 1,119 2.77 Patagonia - 5,877 2.52 48,528 - - 18,030 2,073 3.07 Uuguay - 2,995 1.99 2,709,996 - - 922,357 1,625 2.44 Paaguay 382,390 6,632 10.65 382,409 6,612 10.64 382,402 6,622 10.67 B. Aies - 32,139 3.15 2,950,620 415 0.44 621,317 12,033 3.83 NOA - 1,174 2.17 152,853 - - 69,919 653 2.73 NEA W. - 3,607 2.83 1,462,890 162 0.56 333,719 1,891 3.40 NEA E. - 2,164 2.25 1,058,521 70 0.17 271,882 1,223 2.89 Ente Rios - 14,091 4.20 2,451,317 115 0.29 543,007 3,562 3.76 TOTAL 382,390 79,004 13,232,530 7,527 3,636,406 34,369 (b) Lage Outbeaks (Aveage plus Standad Deviation) Region SO SO-VAC SO-PREV Vaccinated Culled Duation (weeks) Vaccinated Culled Duation (weeks) Vaccinated Culled Duation (weeks) Pampas - 21,591 7.23 997,172 9,274 8.61 4,416,552 432 0.74 Cuyo - 4,926 5.43 248,553 2,987 7.37 517,178 - - Patagonia - 16,281 6.88 50,897 5,750 8.38 120,535 - - Uuguay - 8,485 5.44 2,536,771 4,630 6.76 6,687,413 - - Paaguay - 10,526 7.74-10,624 7.84-10,614 7.78 B. Aies - 86,520 7.96 1,595,690 30,829 9.78 7,310,277 1,123 1.36 NOA - 2,923 5.34 201,602 1,795 7.29 373,115 - - NEA W. - 8,611 6.52 885,262 4,881 8.64 3,619,837 411 1.51 NEA E. - 5,450 5.63 793,074 3,200 7.51 2,612,786 244 0.59 Ente Rios - 38,476 10.63 1,377,877 9,545 9.67 5,965,065 299 0.76 TOTAL - 203,789 8,686,899 83,514 31,622,759 13,123 Region VAC-SO PREV VAC Vaccinated Culled Duation (weeks) Vaccinated Culled Duation (weeks) Vaccinated Culled Duation (weeks) Pampas - 21,136 7.07 4,334,766 412 0.72 971,495 9,214 8.51 Cuyo - 4,822 5.28 506,348 - - 243,115 2,944 7.18 Patagonia - 15,770 6.67 117,916 - - 48,578 5,620 8.19 Uuguay - 8,062 5.28 6,534,518 - - 2,509,720 4,447 6.57 Paaguay 383,099 7,126 11.13 382,975 7,081 11.12 382,752 7,103 11.14 B. Aies - 82,137 7.76 7,117,819 1,077 1.26 1,573,005 31,500 9.69 NOA - 2,985 5.46 368,920 - - 188,023 1,722 7.07 NEA W. - 8,608 6.59 3,534,453 445 1.70 859,750 4,749 8.39 NEA E. - 5,468 5.64 2,548,476 224 0.55 767,102 3,161 7.38 Ente Rios - 36,213 10.39 5,884,788 298 0.74 1,375,283 9,297 9.50 TOTAL 383,099 192,329 31,330,980 9,537 8,918,822 79,756 8.51 Souce: Model Simulations
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 30 Table 3: Shot-un impacts of altenative disease stategies in the Southen Cone Aveage Outbeak Lage Outbeak Stategy Goss Revenue (million USD) Costs (million USD) Net Revenue (million USD) Goss Revenue (million USD) Costs (million USD) Net Revenue (million USD) SO 2,488 28.6 2,460 2,165 68.3 2,097 SO-VAC 2,536 15.6 2,520 2,189 35.8 2,154 SO-PREV 2,740 15.4 2,724 2,686 32.9 2,653 VAC-SO 2,504 26.8 2,477 2,164 64.8 2,100 PREV 2,740 14.4 2,725 2,686 31.4 2,655 VAC 2,536 14.8 2,521 2,189 34.7 2,154 Souce: Model simulations; totals may not add up due to ounding. Table 4: Long-un impacts of altenative disease stategies in the Southen Cone Aveage Outbeak Lage Outbeak Stategy PV of Goss Revenue (million USD) PV of Costs (million USD) PV of Net Revenue (million USD) PV of Goss Revenue (million USD) PV of Costs (million USD) PV of Net Revenue (million USD) SO 19,749 27.2 19,722 19,477 65.0 19,412 SO-VAC 19,161 14.9 19,146 18,873 34.1 18,839 SO-PREV 19,332 14.7 19,317 19,287 31.3 19,256 VAC-SO 19,762 25.5 19,737 19,477 61.7 19,416 PREV 19,332 13.7 19,318 19,287 29.9 19,257 VAC 19,161 14.1 19,146 18,873 33.1 18,840 Souce: Model simulations; totals may not add up due to ounding. PV = Pesent Value. Table 5: Regional diffeences between VAC-SO and SO stategies Aveage Outbeak Lage Outbeak Region VAC-SO Net Benefit (million USD) SO Net Benefit (million USD) Diffeence (million USD) VAC-SO Net Benefit (million USD) SO Net Benefit (million USD) Diffeence (million USD) Patacuyo 585 584 0.52 571 571 0.20 Pampas 13,145 13,137 7.08 12,906 12,903 3.11 Noth 2,206 2,205 1.05 2,178 2,178 (0.45) Uuguay 3,028 3,024 4.96 3,000 3,000 (0.02) Paaguay 773 772 1.20 760 760 0.71 Souce: Model Simulations; totals may not add up due to ounding. The thee egions fo Agentina above ae aggegates of the epidemiological egions (table 2) and ae defined as follows: Patacuyo: Patagonia and Cuyo Pampas: Pampas, Ente Rios, Buenos Aies Noth: NOA, NEA Wes NEA East
DISCOSEM: An Integated Epidemiological-Economic Analysis of Foot and Mouth Disease in the Southen Cone 31 Figue 1 STELLA epesentation of the epidemiological component of DISCOSEM