The Feasibility of Indemnification and Check-off Funded Programs to Manage Invasive Species Risks in Agriculture Report of Progress Barry K. Goodwin Nicholas E. Piggott North Carolina State University Overview Increased integration of world markets & international mobility of goods and people • heightened concerns regarding harmful invasive species Threat is substantial to agriculture & some significant damages have already been realized in U.S. agriculture Overview… Current response has been to provide ad hoc disaster assistance targeted to specific commodities and/or regions An alternative strategy might involve either a fund or insurance program to protect producers from risks associated with specific invasive specifies Objectives: Two Fold Evaluate economic issues in design of voluntary insurance and mandatory checkoff programs Statistical modeling of the risks associated with several case studies with the aim of pricing insurance or determining optimal check-off contribution rates Case studies: Asiatic canker, karnal bunt, and soybean rust Objectives…. Model contamination risks and expected losses at county level for a representative producer Attention to exogenous factors associated with transmission including Trade and transportation patterns Migrant labor and harvesting crews Important characteristics of land and farms Weather Statistical models explicitly measure spatial patterns of risks and transmission Uncertainty of the Risk • Harmful economic effects of an infestation or contamination by a NIS are similar to effects of any other pest or disease –Exposure may lead to yield losses or affect quality (crops and livestock) • What is different is the uncertainty associated with the risks from many different known and unknown pests and diseases –Potential for catastrophic losses from new NIS may far exceed losses from more common pests and diseases State of Affairs • Congress typically responds to losses from outbreaks of pests and disease by providing direct ad hoc payment to affected producers –E.g. $32 m on karnal bunt in wheat in 1990s –Some states also provide financial assistance • Costs of prevention and managing spread of diseases and outbreak impose large cost on govt “if the CCC fund cannot be accessed, then we need to consider the development of an invasive pest trust fund” (Carl Loop, President FL Farm Bur., congressional testimony (Jan 2000) Indemnification Programs • Feasibility and operational aspects of a program for indemnifying producers against specific perils –current programs may not be entirely comprehensive and sufficient • e.g, Citrus canker can involve multi-years of loss in a grove; Indemnification Programs… • Two options are considered: –Mandatory program that operates using a “check-off” on production, all producers pay into a fund used to cover losses • might also fund prevention/eradication programs –Voluntary indemnification program that measures risks relevant to a specific threat and provides coverage • “additional insurance” beyond what is already available Usefulness of this Work • Should be of interest to state and federal policymakers currently faced with developing ways to manage these risks • An indemnification involving insurance or checkoff could be independent of government support or partially subsidized • These “self-help” alternatives recognized that some of the risk should be internalized (or borne) by those who have the most to lose and not entirely borne by the taxpayer Premium or Check-off Rate Needed to Cover Expected Losses? • Under both scenarios the key parameter is the appropriate premium or check-off rate that will cover expected losses • We are developing and evaluating methods of measuring the risks associated with these losses • Analogous to deriving measures of the actuarially fair insurance rate that would be needed to operate a specific peril program Research Methods • Measuring the risk requires measuring the probability density underlying risks (e.g., yield losses due to the specific peril under consideration) Prob. Yield Indemnities & Costs • For a program that reimburses producers for yields (y) that are beneath a certain proportion (l) of their expected (mean) yield (m) Indemnities = p. (max{ (l m - y), 0}) p the price at which losses are compensated Premium or Check-off Rate • Insurance program or check-off requires a premium or mandated contribution rate determined by expected payouts • For p=1, expected loss is given by E(L)=Prob(y<lm)*[lm-E(y|y<lm)] • Define F() and f() to be the cumulative probability distribution functions (cdf) and the probability density function (pdf) Premium or Check-off Rate • E(Loss) = Pr(loss)*(Loss|Loss Occurs) • Define F() and f() to be the cumulative probability distribution functions (cdf) and the probability density function (pdf) and the premium or check of rate (R) can be shown to be equal to lm f ( y ) ydy lm E ( Loss) f ( y )dy lm 0 lm 0 0 f ( y)dy lm f ( y ) ydy 0 F (lm ) lm F (lm ) R E ( Loss) lm Economics of Check-off Funded Prevention/eradication Program Price S’ a P’’ P’ P S+T E’’ b c d f e g l S E’ h i m n k p j o E Gains in Total Economic Surplus r q s D t D’ Q’ ’ Q’ Q Quantity Challenging Modeling Questions • What is the appropriate form of the distribution [f(y)]? • Are parametric densities appropriate or less restrictive techniques preferred? • What factors should the distribution be conditioned on? • What are the spatial-temporal relationships associated with the invasive species? Case Study: Citrus Canker in FL • Large concern to Florida citrus • Threat since 1910s but outbreak in residential citrus trees in Dade county in Sept of 1995 – Triggered widespread quarantines and mandatory destruction of citrus stocks – Over 1.3 million trees destroyed • Spatio-temporal aspects of transmission especially interesting for evaluating risk Source: Gottwald et. al. (2001) Spatio-Temporal Impacts of yi,t f(yz) i=9 yz t=1 t=2 t=3 t=4 Model to Estimate the Density Functions in a Localized Area County Probability Density Functions f(yi ) f ( yi ,t ) f ( yi ,t yi ,t k , y j ,t , y j ,t k , θ) + i ,t where i 1, 2, ........N (counties) t 1, 2, ........T (periods) θ vector of exogenous factors i ,t error term Modeling Issues • Spatio-temporal correlation: –Contamination spreads along avenues characterized and influenced by spatial and temporal characteristics –Common analysis in epidemiological studies • The process is likely to be characterized by models with large numbers of parameters • We will address this using hierarchical models which consider stages or layers of relationships Modeling Issues…. • We likely will adopt a mixture distribution model that recognizes different states of nature (e.g., recent or nearby infection versus no obvious infections) • Similar models have been applied to model catastrophic versus “normal” crop risks (Goodwin and Ker) • Our analysis will be conducted in a Bayesian context using diffuse priors and the model fitting criterion proposed by Gelfand and Ghosh Modeling Issues…. • Account for the catastrophic risk associated with spread through a significant hurricane event (a significant catalyst for infection spread) • This has not been observed in our data and thus we must assume “tail” probabilities that are associated with events that are nontrivial in probability but that have not been experienced –historical hurricane records (data that we have already assembled) Florida Data • We traveled to Florida in April to learn about citrus canker (transmission, damaged, current eradication program etc) • Met with representatives of citrus canker eradication program, USDA, and Florida Dept. of Ag. (Tim Gottwald, Tom Gates, Fritz Roka) • Recently received authorization (lengthy process) • On July 29 we received an initial dataset of commercial grove data (inspection results at the sub-grove unit level) Other Applications • Methodology and techniques being developed are general and can be applied to other spatio-temporal risks of invasive species –e.g. spread of soybean rust rising concern in North Carolina with new import facility in Wilmington • We may shift focus of second application to rust rather than karnal bunt as we were notified that the “karnal bunt pest risk assessments (basis for data) are still in draft format and not ready to be released”